{"id":13444376,"url":"https://github.com/HCIILAB/Scene-Text-Recognition","last_synced_at":"2025-03-20T18:32:23.286Z","repository":{"id":45839537,"uuid":"186785087","full_name":"HCIILAB/Scene-Text-Recognition","owner":"HCIILAB","description":null,"archived":false,"fork":false,"pushed_at":"2021-12-17T09:05:40.000Z","size":1106,"stargazers_count":603,"open_issues_count":2,"forks_count":117,"subscribers_count":32,"default_branch":"master","last_synced_at":"2024-08-01T04:02:07.580Z","etag":null,"topics":[],"latest_commit_sha":null,"homepage":null,"language":null,"has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":null,"status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/HCIILAB.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":null,"code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null}},"created_at":"2019-05-15T08:38:44.000Z","updated_at":"2024-07-27T07:54:39.000Z","dependencies_parsed_at":"2022-07-17T14:47:03.629Z","dependency_job_id":null,"html_url":"https://github.com/HCIILAB/Scene-Text-Recognition","commit_stats":null,"previous_names":[],"tags_count":0,"template":false,"template_full_name":null,"repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/HCIILAB%2FScene-Text-Recognition","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/HCIILAB%2FScene-Text-Recognition/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/HCIILAB%2FScene-Text-Recognition/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/HCIILAB%2FScene-Text-Recognition/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/HCIILAB","download_url":"https://codeload.github.com/HCIILAB/Scene-Text-Recognition/tar.gz/refs/heads/master","host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":221792892,"owners_count":16881289,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2022-07-04T15:15:14.044Z","host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":[],"created_at":"2024-07-31T04:00:21.309Z","updated_at":"2024-10-28T06:30:44.327Z","avatar_url":"https://github.com/HCIILAB.png","language":null,"funding_links":[],"categories":["Uncategorized","References","Other lists"],"sub_categories":["Uncategorized"],"readme":"# Scene Text Recognition Resources\n[\u003cp align='right'\u003eAuthor: 陈晓雪\u003c/p\u003e](https://github.com/CCchenxiaoxue)\nThe paper \"Text Recognition in the Wild: A Survey\" (accepted to appear in ACM Computing Surveys) in [arXiv](https://arxiv.org/pdf/2005.03492v3.pdf) version is available now.\n\n# ❗❗ Newest Version Can be Found Here ❗❗\n## This repository is no longer maintaining now, and you can refer to our newest one\n## [Scene Text Recognition Recommendations](https://github.com/HCIILAB/Scene-Text-Recognition-Recommendations)\n\n## Updates\n\nDec 24, 2019: add 20 papers and update corresponding tables. \n\nFeb 29, 2020: add AAAI-2020 papers and update corresponding tables. \n\nMay 8, 2020: add CVPR-2020 papers and update corresponding tables. \n\nDec 8, 2020: add 11 papers and update corresponding tables. You can download the new [Excel](https://pan.baidu.com/s/1xitxu7R5hw27pVV7eJ1c7w) prepared by us. (Password: sj2t)\n\n***\n\n\u003c!-- MarkdownTOC --\u003e\n\n- [1. Datasets](#1-datasets)\n  - [1.1 Regular Latin Datasets](#11-regular-latin-datasets)\n  - [1.2 Irregular Latin Datasets](#12-irregular-latin-datasets)\n  - [1.3 Multilingual Datasets](#13-multilingual-datasets)\n  - [1.4 Synthetic Datasets](#14-synthetic-datasets)\n  - [1.5 Comparison of the Benchmark Datasets](#15-comparison-of-the-benchmark-datasets)\n- [2. Performance Comparison of Recognition Algorithms](#2-performance-comparison-of-recognition-algorithms)\n  - [2.1 Characteristics Comparison of Recognition Approaches](#21-characteristics-comparison-of-recognition-approaches)\n  - [2.2 Performance Comparison on Benchmark Datasets](#22-performance-comparison-on-benchmark-datasets)\n       - [2.2.1 Performance Comparison of Recognition Algorithms on Regular Latin Datasets](#221-performance-comparison-of-recognition-algorithms-on-regular-latin-datasets)\n       - [2.2.2 Performance Comparison of Recognition Algorithms on Irregular Latin Datasets](#222-performance-comparison-of-recognition-algorithms-on-irregular-latin-datasets)\n- [3. Survey](#3-survey)\n- [4. OCR Service](#4-ocr-service)\n- [5. References](#5-references)\n- [6.Help](#6help)\n- [7.Copyright](#7copyright)\n\n\u003c!-- /MarkdownTOC --\u003e\n\n\u003ca id=\"1-datasets\"\u003e\u003c/a\u003e\n## 1. Datasets\n\n\u003ca id=\"11-regular-latin-datasets\"\u003e\u003c/a\u003e\n### 1.1 Regular Latin Datasets\n\n- IIIT5K[31]：\n  * **Introduction:**  The IIIT5K dataset [31] contains 5,000 text instance images: 2,000 for training and 3,000 for testing. It contains words from street scenes and from originally-digital images. Every image is associated with a 50 -word lexicon and a 1,000 -word lexicon. Specifically, the lexicon consists of a ground-truth word and some randomly picked words.\n  * **Link:** [IIIT5K-download](http://cvit.iiit.ac.in/research/projects/cvit-projects/the-iiit-5k-word-dataset)\n- SVT[1]：\n  * **Introduction:**  The SVT dataset [1] contains 350 images: 100 for training and 250 for testing. Some images are severely corrupted by noise, blur, and low resolution. Each image is associated with a 50 -word lexicon.\n  * **Link:** [SVT-download](http://vision.ucsd.edu/~kai/svt/)\n- ICDAR 2003(IC03)[33]：\n  * **Introduction:**  The IC03 dataset [33] contains 509 images: 258 for training and 251 for testing. Specifically, it contains 867 cropped text instances after discarding images that contain non-alphanumeric characters or less than three characters. Every image is associated with a 50 -word lexicon and a full-word lexicon. Moreover, the full lexicon combines all lexicon words.\n  * **Link:** [IC03-download](http://www.iapr-tc11.org/mediawiki/index.php?title=ICDAR_2003_Robust_Reading_Competitions)\n- ICDAR 2013(IC13)[34]：\n  * **Introduction:**  The IC13 dataset [34] contains 561 images: 420 for training and 141 for testing. It inherits data from the IC03 dataset and extends it with new images. Similar to IC03 dataset, the IC13 dataset contains 1,015 cropped text instance images after removing the words with non-alphanumeric characters. No lexicon is associated with IC13 . Notably, 215 duplicate text instance images [65] exist between the IC03 training dataset and the IC13 testing dataset. Therefore, care should be taken regarding the overlapping data when evaluating a model on the IC13 testing data. \n  * **Link:** [IC13-download](http://dagdata.cvc.uab.es/icdar2013competition/?ch=2\u0026com=downloads)\n- SVHN[45]：\n  * **Introduction:** The SVHN [45] dataset contains more than 600,000 digits of house numbers in natural scenes. It is obtained from a large number of street view images using a combination of automated algorithms and the Amazon Mechanical Turk (AMT) framework. The SVHN dataset was typically  used for scene digit recognition.  \n  * **Link:** [SVHN-download](http://ufldl.stanford.edu/housenumbers/)\n\n\u003ca id=\"12-irregular-latin-datasets\"\u003e\u003c/a\u003e\n### 1.2 Irregular Latin Datasets\n\n- SVT-P[35]：\n  - **Introduction:** The SVT-P [35] dataset contains 238 images with 639 cropped text instances. It is specifically designed to evaluate perspective distorted text recognition. It is built based on the original SVT dataset by selecting the images at the same address on Google Street View but with different view angles. Therefore, most text instances are heavily distorted by the non-frontal view angle. Moreover, each image is associated with a 50-word lexicon and a full-word lexicon.  \n  - **Link:** [SVT-P-download](https://pan.baidu.com/s/1rhYUn1mIo8OZQEGUZ9Nmrg )  \\(Password : vnis)\n- CUTE80[36]：\n  - **Introduction:**  The CUTE80 dataset [36] contains 80 high-resolution images with 288 cropped text\n    instances. It focuses on curved text recognition. Most images in CUTE80 have a complex background, perspective distortion, and poor resolution. No lexicon is associated with CUTE80.\n  - **Link:** [CUTE80-download](http://cs-chan.com/downloads_CUTE80_dataset.html)\n- ICDAR 2015(IC15)[37]：\n  - **Introduction:** The IC15 dataset [37] contains 1,500 images: 1,000 for training and 500 for testing.\n    Specifically, it contains 2,077 cropped text instances, including more than 200 irregular text samples. As text images were taken by Google Glasses without ensuring the image quality, most of the text is very small, blurred, and multi-oriented. No lexicon is provided.\n  - **Link:** [IC15-download](http://rrc.cvc.uab.es/?ch=4\u0026com=downloads)\n- COCO-Text[38]：\n  - **Introduction:** The COCO-Text dataset [38] contains 63,686 images with 145,859 cropped text instances. It is the first large-scale dataset for text in natural images and also the first dataset to annotate scene text with attributes such as legibility and type of text. However, no lexicon is associated with COCO-Text.  \n  - **Link:** [COCO-Text-download](https://vision.cornell.edu/se3/coco-text-2/)\n- Total-Text[39]：\n  - **Introduction:** The Total-Text dataset [39] contains 1,555 images with 11,459 cropped text instance images. It focuses on curved scene text recognition. Images in Total-Text have more than three different orientations, including horizontal, multi-oriented, and curved. No lexicon is associated with Total-Text.\n  - **Link:** [Total-Text-download](https://github.com/cs-chan/Total-Text-Dataset)\n\n\u003ca id=\"13-multilingual-datasets\"\u003e\u003c/a\u003e\n### 1.3 Multilingual Datasets\n\n- RCTW-17(RCTW competition，ICDAR17)[40]：\n  - **Introduction:** The RCTW-17 dataset contains 12,514 images: 11,514 for training and 1,000 for testing. Most are natural images collected by cameras or mobile phones, whereas others are digital-born. Text instances are annotated with labels, fonts, languages, etc.    \n  - **Link:** [RCTW-17-download](http://rctw.vlrlab.net/dataset/)\n- MTWI(competition)[41]：\n  - **Introduction:** The MTWI dataset contains 20,000 images. This is the first dataset constructed by Chinese and Latin web text. Most images in MTWI have a relatively high resolution and cover diverse types of web text, including multi-oriented text, tightly-stacked text, and complex-shaped text.  \n  - **Link:** [MTWI-download ](https://pan.baidu.com/s/1SUODaOzV7YOPkrun0xSz6A#list/path=%2F)  \\(Password:gox9)\n- CTW[42]：\n  - **Introduction:** The CTW dataset includes 32,285 high-resolution street view images with 1,018,402 character instances. All images have character-level annotations: the underlying character, the bounding box, and six other attributes.\n  - **Link:** [CTW-download](https://ctwdataset.github.io/)\n- SCUT-CTW1500[43]：\n  - **Introduction:** The SCUT-CTW1500 dataset contains 1,500 images: 1,000 for training and 500\n    for testing. In particular, it provides 10,751 cropped text instance images, including 3,530 with curved text. The images are manually harvested from the Internet, image libraries such as Google Open-Image, or phone cameras. The dataset contains a lot of horizontal and multi-oriented text   \n  - **Link:** [SCUT-CTW1500-download](https://github.com/Yuliang-Liu/Curve-Text-Detector)\n* LSVT(LSVT competition, ICDAR2019)[57]:\n  * **Introduction:** The LSVT dataset contains 20,000 testing samples, 30,000 fully annotated training samples, and 400,000 training samples with weak annotations (i.e., with partial labels). All images are captured from streets and reflect a large variety of complicated real-world scenarios, e.g., store fronts and landmarks.  \n  * **Link:** [LSVT-download](https://rrc.cvc.uab.es/?ch=16\u0026com=downloads)\n* ArT(ArT competition, ICDAR2019)[58]:\n  * **Introduction:** The ArT dataset [58] contains 10,166 images: 5,603 for training and 4,563 for testing. ArT is a combination of Total-Text, SCUT-CTW 1500 , and Baidu Curved Scene Text 4 , which was collected to introduce the arbitrary-shaped text problem. Moreover, all existing text shapes (i.e., horizontal, multi-oriented, and curved) have multiple occurrences in the ArT dataset.\n  * **Link:** [ArT-download](https://rrc.cvc.uab.es/?ch=16\u0026com=downloads)\n* ReCTS-25k(ReCTS competition, ICDAR2019)[59]:\n  * **Introduction:** The ReCTS-25k dataset [59] contains 25,000 images: 20,000 for training and 5,000 for testing.  All the images are from the Meituan-Dianping Group, collected by Meituan business mer-\n    chants, using phone cameras under uncontrolled conditions. Specifically, ReCTS-25 k dataset mainly contains images of Chinese text on signboards.\n  * **Link:** [ReCTS-download](https://rrc.cvc.uab.es/?ch=16\u0026com=downloads)\n* MLT(MLTcompetition, ICDAR2019) [81]:\n  * **Introduction:** The MLT-2019 dataset [81] contains 20,000 images: 10,000 for training (1,000 per language) and 10,000 for testing. The dataset includes ten languages, representing seven different scripts: Arabic, Bangla, Chinese, Devanagari, English, French, German, Italian, Japanese, and Korean. The number of images per script is equal.  \n  * **Link:** [MLT-download](https://rrc.cvc.uab.es/?ch=15\u0026com=downloads)  \n\n\u003ca id=\"14-synthetic-datasets\"\u003e\u003c/a\u003e\n### 1.4 Synthetic Datasets\n\n* Synth90k [53] : \n  * **Introduction:** The Synth90k dataset contains 9 million synthetic text instance images from a set of 90k common English words. Words are rendered onto natural images with random transformations and effects, such as random fonts, colors, blur, and noises. Synth90k dataset can emulate the distribution of scene text images and can be used instead of real-world data to train data-hungry deep learning algorithms. Besides, every image is annotated with a ground-truth word.  \n  * **Link:** [Synth90k-download](http://www.robots.ox.ac.uk/~vgg/data/text/)\n* SynthText [54] : \n  * **Introduction:** The SynthText dataset contains 800,000 images with 6 million synthetic text instances. As in the generation of Synth90k dataset, the text sample is rendered using a randomly selected font and transformed according to the local surface orientation. Moreover, each image is annotated with a ground-truth word.  \n  * **Link:** [SynthText-download](https://github.com/ankush-me/SynthText)\n* Verisimilar Synthesis [73] : \n  * **Introduction:** The Verisimilar Synthesis dataset [73] contains 5 million synthetic text instance images. Given background images and source texts, a semantic map and a saliency map are first\n    determined which are then combined to identify semantically sensible and apt locations for text embedding. The color, brightness, and orientation of the source texts are further determined adaptively according to the color, brightness, and contextual structures around the embedding locations within the background image.  \n  * **Link:** [Verisimilar Synthesis](https://github.com/fnzhan/Verisimilar-Image-Synthesis-for-Accurate-Detection-and-Recognition-of-Texts-in-Scenes)\n* UnrealText [88]:\n  * **Introduction:** The UnrealText dataset [88] contains 600K synthetic images with 12 million cropped text instances. It is developed upon Unreal Engine 4 and the UnrealCV plugin [89]. Text instances are regarded as planar polygon meshes with text foregrounds loaded as texture. These meshes are placed in suitable positions in 3D world, and rendered together with the scene as a whole. The same font set from [Google Fonts](https://fonts.google.com/) and the same text corpus, i.e., Newsgroup20, are used as SynthText does.    \n  * **Link:** [Verisimilar Synthesis](https://github.com/fnzhan/Verisimilar-Image-Synthesis-for-Accurate-Detection-and-Recognition-of-Texts-in-Scenes)\n\n\u003ca id=\"15-comparison-of-the-benchmark-datasets\"\u003e\u003c/a\u003e\n### 1.5 Comparison of the Benchmark Datasets\n\n\u003ctable cellspacing=\"0\" border=\"0\"\u003e\n  \u003ccolgroup width=\"271\"\u003e\u003c/colgroup\u003e\n  \u003ccolgroup width=\"179\"\u003e\u003c/colgroup\u003e\n  \u003ccolgroup width=\"89\"\u003e\u003c/colgroup\u003e\n  \u003ccolgroup width=\"122\"\u003e\u003c/colgroup\u003e\n  \u003ccolgroup width=\"127\"\u003e\u003c/colgroup\u003e\n  \u003ccolgroup width=\"89\"\u003e\u003c/colgroup\u003e\n  \u003ccolgroup width=\"179\"\u003e\u003c/colgroup\u003e\n  \u003ccolgroup width=\"177\"\u003e\u003c/colgroup\u003e\n  \u003ccolgroup span=\"7\" width=\"89\"\u003e\u003c/colgroup\u003e\n  \u003ctr\u003e\n    \u003ctd colspan=15 height=\"34\" align=\"center\"\u003e\u003cb\u003eComparison of the Benchmark Datasets\u003c/b\u003e\u003c/td\u003e\n    \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd rowspan=2 height=\"39\" align=\"center\"\u003e\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;Datasets\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u003c/td\u003e\n    \u003ctd rowspan=2 align=\"center\"\u003eLanguage\u003c/td\u003e\n    \u003ctd colspan=6 align=\"center\"\u003eImages\u003c/td\u003e\n    \u003ctd colspan=4 align=\"center\"\u003eLexicon\u003c/td\u003e\n    \u003ctd colspan=2 align=\"center\"\u003eLabel\u003c/td\u003e\n    \u003ctd rowspan=2 align=\"center\"\u003eType\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd align=\"center\"\u003ePictures\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eTraining Pictures\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eTesting Pictures\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eInstances\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eTraining Instances\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eTesting Instances\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"50\" sdnum=\"2052;\"\u003e50\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e1k\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eFull\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eNone\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eChar\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eWord\u003c/td\u003e\n    \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eIIIT5K[31]\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eEnglish\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"1120\" sdnum=\"2052;\"\u003e1120\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"380\" sdnum=\"2052;\"\u003e380\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"740\" sdnum=\"2052;\"\u003e740\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"5000\" sdnum=\"2052;\"\u003e5000\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2000\" sdnum=\"2052;\"\u003e2000\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"3000\" sdnum=\"2052;\"\u003e3000\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e ×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eRegular\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eSVT[32]\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eEnglish\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"350\" sdnum=\"2052;\"\u003e350\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"100\" sdnum=\"2052;\"\u003e100\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"250\" sdnum=\"2052;\"\u003e250\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"725\" sdnum=\"2052;\"\u003e725\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"211\" sdnum=\"2052;\"\u003e211\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"514\" sdnum=\"2052;\"\u003e514\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e ×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e ×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e ×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eRegular\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eIC03[33]\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eEnglish\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"509\" sdnum=\"2052;\"\u003e509\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"258\" sdnum=\"2052;\"\u003e258\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"251\" sdnum=\"2052;\"\u003e251\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2268\" sdnum=\"2052;\"\u003e2268\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"1157\" sdnum=\"2052;\"\u003e1157\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"1111\" sdnum=\"2052;\"\u003e1111\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eRegular\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eIC13[34]\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eEnglish\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"561\" sdnum=\"2052;\"\u003e561\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"420\" sdnum=\"2052;\"\u003e420\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"141\" sdnum=\"2052;\"\u003e141\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"5003\" sdnum=\"2052;\"\u003e5003\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"3564\" sdnum=\"2052;\"\u003e3564\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"1439\" sdnum=\"2052;\"\u003e1439\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e ×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e ×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e ×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eRegular\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eSVHN[45]\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eDigits\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"600000\" sdnum=\"2052;\"\u003e600000\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"573968\" sdnum=\"2052;\"\u003e573968\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"26032\" sdnum=\"2052;\"\u003e26032\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"600000\" sdnum=\"2052;\"\u003e600000\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"573968\" sdnum=\"2052;\"\u003e573968\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"26032\" sdnum=\"2052;\"\u003e26032\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e ×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e ×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e ×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eRegular\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eSVT-P[35]\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eEnglish\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"238\" sdnum=\"2052;\"\u003e238\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"0\" sdnum=\"2052;\"\u003e0\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"238\" sdnum=\"2052;\"\u003e238\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"639\" sdnum=\"2052;\"\u003e639\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"0\" sdnum=\"2052;\"\u003e0\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"639\" sdnum=\"2052;\"\u003e639\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e ×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e ×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eIrregular\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eCUTE80[36]\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eEnglish\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"80\" sdnum=\"2052;\"\u003e80\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"0\" sdnum=\"2052;\"\u003e0\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"80\" sdnum=\"2052;\"\u003e80\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"288\" sdnum=\"2052;\"\u003e288\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"0\" sdnum=\"2052;\"\u003e0\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"288\" sdnum=\"2052;\"\u003e288\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e ×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e ×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e ×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e ×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eIrregular\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eIC15[37]\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eEnglish\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"1500\" sdnum=\"2052;\"\u003e1500\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"1000\" sdnum=\"2052;\"\u003e1000\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"500\" sdnum=\"2052;\"\u003e500\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"6545\" sdnum=\"2052;\"\u003e6545\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"4468\" sdnum=\"2052;\"\u003e4468\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2077\" sdnum=\"2052;\"\u003e2077\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e ×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e ×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e ×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e ×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eIrregular\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eCOCO-Text[38]\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eEnglish\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"63686\" sdnum=\"2052;\"\u003e63686\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"43686\" sdnum=\"2052;\"\u003e43686\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"10000\" sdnum=\"2052;\"\u003e10000\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"145859\" sdnum=\"2052;\"\u003e145859\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"118309\" sdnum=\"2052;\"\u003e118309\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"27550\" sdnum=\"2052;\"\u003e27550\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e ×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e ×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e ×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e ×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eIrregular\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eTotal-Text[39]\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eEnglish\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"1555\" sdnum=\"2052;\"\u003e1555\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"1255\" sdnum=\"2052;\"\u003e1255\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"300\" sdnum=\"2052;\"\u003e300\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"11459\" sdnum=\"2052;\"\u003e11459\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"11166\" sdnum=\"2052;\"\u003e11166\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"293\" sdnum=\"2052;\"\u003e293\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e ×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e ×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e ×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e ×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eIrregular\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eRCTW-17[40]\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eChinese/English\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"12514\" sdnum=\"2052;\"\u003e12514\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"11514\" sdnum=\"2052;\"\u003e11514\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"1000\" sdnum=\"2052;\"\u003e1000\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e ×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e ×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e ×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e ×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eRegular\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eMTWI[41]\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eChinese/English\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"20000\" sdnum=\"2052;\"\u003e20000\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"10000\" sdnum=\"2052;\"\u003e10000\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"10000\" sdnum=\"2052;\"\u003e10000\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"290206\" sdnum=\"2052;\"\u003e290206\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"141476\" sdnum=\"2052;\"\u003e141476\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"148730\" sdnum=\"2052;\"\u003e148730\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e ×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e ×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e ×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e ×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eRegular\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eCTW[42]\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eChinese/English\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"32285\" sdnum=\"2052;\"\u003e32285\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"25887\" sdnum=\"2052;\"\u003e25887\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"3269\" sdnum=\"2052;\"\u003e3269\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"1018402\" sdnum=\"2052;\"\u003e1018402\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"812872\" sdnum=\"2052;\"\u003e812872\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"103519\" sdnum=\"2052;\"\u003e103519\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e ×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e ×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e ×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eRegular\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eSCUT-CTW1500[43]\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eChinese/English\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"1500\" sdnum=\"2052;\"\u003e1500\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"1000\" sdnum=\"2052;\"\u003e1000\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"500\" sdnum=\"2052;\"\u003e500\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"10751\" sdnum=\"2052;\"\u003e10751\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"7683\" sdnum=\"2052;\"\u003e7683\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"3068\" sdnum=\"2052;\"\u003e3068\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e ×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e ×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e ×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e ×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eIrregular\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eLSVT[57], [63]\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eChinese/English\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"450000\" sdnum=\"2052;\"\u003e450000\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"30000\" sdnum=\"2052;\"\u003e30000\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"20000\" sdnum=\"2052;\"\u003e20000\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e ×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e ×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e ×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e ×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eIrregular\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eArT[58]\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eChinese/English\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"10166\" sdnum=\"2052;\"\u003e10166\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"5603\" sdnum=\"2052;\"\u003e5603\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"4563\" sdnum=\"2052;\"\u003e4563\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"98455\" sdnum=\"2052;\"\u003e98455\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"50029\" sdnum=\"2052;\"\u003e50029\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"48426\" sdnum=\"2052;\"\u003e48426\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e ×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e ×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e ×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e ×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eIrregular\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eReCTS-25k[59]\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eChinese/English\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"25000\" sdnum=\"2052;\"\u003e25000\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"20000\" sdnum=\"2052;\"\u003e20000\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"5000\" sdnum=\"2052;\"\u003e5000\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"119713\" sdnum=\"2052;\"\u003e119713\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"108924\" sdnum=\"2052;\"\u003e108924\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"10789\" sdnum=\"2052;\"\u003e10789\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e ×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e ×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e ×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eIrregular\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eMLT[81]\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eMultilingual\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"20000\" sdnum=\"2052;\"\u003e20000\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"10000\" sdnum=\"2052;\"\u003e10000\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"10000\" sdnum=\"2052;\"\u003e10000\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"191639\" sdnum=\"2052;\"\u003e191639\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"89177\" sdnum=\"2052;\"\u003e89177\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"102462\" sdnum=\"2052;\"\u003e102462\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e ×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e ×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e ×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e ×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eIrregular\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eSynth90k[53]\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eEnglish\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e~9000000\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e~9000000\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e ×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e ×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e ×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e ×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eRegular\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eSynthText[54]\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eEnglish\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e~6000000\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e~6000000\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e ×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e ×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e ×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eRegular\u003c/td\u003e\n  \u003c/tr\u003e\n    \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eVerisimilar Synthesis[73]\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eEnglish\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e~5000000\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e ×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e ×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e ×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e ×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eRegular\u003c/td\u003e\n  \u003c/tr\u003e\n    \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eUnrealText[88]\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eEnglish\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e~600000\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e~12000000\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e ×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e ×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e ×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e √\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eRegular\u003c/td\u003e\n  \u003c/tr\u003e\n\u003c/table\u003e\n\n***\n\n\u003ca id=\"2-performance-comparison-of-recognition-algorithms\"\u003e\u003c/a\u003e\n## 2. Performance Comparison of Recognition Algorithms\n\n\u003ca id=\"21-characteristics-comparison-of-recognition-approaches\"\u003e\u003c/a\u003e\n### 2.1 Characteristics Comparison of Recognition Approaches\n\nIt is notable that 1) \"Reg\" stands for regular Latin datasets. 2) \"Irreg\" stands for irregular Latin datasets. 3) \"Seg\" denotes the segmentation-based methods. 4) \"Extra\" indicates the methods that use the extra datasets other than Synth90k and SynthText. 5) \"CTC\" represents the methods that apply the CTC-based algorithm to decode. 6) \"Attn\" represents the method that apply the attention mechanism to decode.\n\nYou can also download the new [Excel](https://pan.baidu.com/s/1xitxu7R5hw27pVV7eJ1c7w) prepared by us. (Password: sj2t)\n\n\u003ctable cellspacing=\"0\" border=\"0\"\u003e\n  \u003ccolgroup width=\"271\"\u003e\u003c/colgroup\u003e\n  \u003ccolgroup span=\"3\" width=\"89\"\u003e\u003c/colgroup\u003e\n  \u003ccolgroup width=\"104\"\u003e\u003c/colgroup\u003e\n  \u003ccolgroup span=\"3\" width=\"89\"\u003e\u003c/colgroup\u003e\n  \u003ccolgroup span=\"2\" width=\"86\"\u003e\u003c/colgroup\u003e\n  \u003ccolgroup width=\"832\"\u003e\u003c/colgroup\u003e\n  \u003ctr\u003e\n    \u003ctd colspan=11 height=\"34\" align=\"center\"\u003e\u003cb\u003eCharacteristics Comparison of Recognition Approaches\u003c/b\u003e\u003c/td\u003e\n    \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd rowspan=2 height=\"39\" align=\"center\"\u003e\u003cb\u003e\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;Method\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u003c/b\u003e\u003c/td\u003e\n    \u003ctd rowspan=2 align=\"center\"\u003e\u003cb\u003eCode\u003c/b\u003e\u003c/td\u003e\n    \u003ctd rowspan=2 align=\"center\"\u003e\u003cb\u003eRegular\u003c/b\u003e\u003c/td\u003e\n    \u003ctd rowspan=2 align=\"center\"\u003e\u003cb\u003eIrregular\u003c/b\u003e\u003c/td\u003e\n    \u003ctd rowspan=2 align=\"center\"\u003e\u003cb\u003eSegmentation\u003c/b\u003e\u003c/td\u003e\n    \u003ctd rowspan=2 align=\"center\"\u003e\u003cb\u003eExtra data\u003c/b\u003e\u003c/td\u003e\n    \u003ctd rowspan=2 align=\"center\"\u003e\u003cb\u003eCTC\u003c/b\u003e\u003c/td\u003e\n    \u003ctd rowspan=2 align=\"center\"\u003e\u003cb\u003eAttention\u003c/b\u003e\u003c/td\u003e\n    \u003ctd rowspan=2 align=\"center\"\u003e\u003cb\u003eSource\u003c/b\u003e\u003c/td\u003e\n    \u003ctd rowspan=2 align=\"center\"\u003e\u003cb\u003eTime\u003c/b\u003e\u003c/td\u003e\n    \u003ctd rowspan=2 align=\"center\"\u003e\u003cb\u003e\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;Highlight\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u003c/b\u003e\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eWang et al. [1] : ABBYY\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e ×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eICCV\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2011\" sdnum=\"2052;\"\u003e2011\u003c/td\u003e\n    \u003ctd align=\"center\"\u003ea state-of-the-art text detector + a leading commercial OCR engine\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eWang et al. [1] : SYNTH+PLEX\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eICCV\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2011\" sdnum=\"2052;\"\u003e2011\u003c/td\u003e\n    \u003ctd align=\"center\"\u003ethe baseline of scene text recognition\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eMishra et al. [2]\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eBMVC\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2012\" sdnum=\"2052;\"\u003e2012\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e1) incorporating higher order statistical language models to recognize words in an unconstrained manner 2) introducing IIIT5K-word dataset\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eWang et al. [3]\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eICPR\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2012\" sdnum=\"2052;\"\u003e2012\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eCNNs + Non-maximal suppression + beam search\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eGoel et al. [4] : wDTW\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eICDAR\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2013\" sdnum=\"2052;\"\u003e2013\u003c/td\u003e\n    \u003ctd align=\"center\"\u003erecognizing text by matching the scene and synthetic image features with wDTW\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eBissacco et al. [5] : PhotoOCR\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eICCV\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2013\" sdnum=\"2052;\"\u003e2013\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eapplying a network with five hidden layers for character classification\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003ePhan et al. [6]\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eICCV\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2013\" sdnum=\"2052;\"\u003e2013\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e1) MSER + SIFT descriptors + SVM 2) introducing the SVT-P datasets\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eAlsharif et al. [7] : HMM/Maxout\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eICLR\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2014\" sdnum=\"2052;\"\u003e2014\u003c/td\u003e\n    \u003ctd align=\"center\"\u003econvolutional Maxout networks + Hybrid HMM\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eAlmazan et al [8] : KCSR\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eTPAMI\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2014\" sdnum=\"2052;\"\u003e2014\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eembedding word images and text strings in a common vectorial subspace and interpreting the task of recognition and retrieval as a nearest neighbor problem\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eYao et al. [9] : Strokelets\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eCVPR\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2014\" sdnum=\"2052;\"\u003e2014\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eproposing a novel multi-scale representation for scene text recognition: strokelets\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eR.-Serrano et al.[10] : Label embedding\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eIJCV\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2015\" sdnum=\"2052;\"\u003e2015\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eembedding word labels and word images into a common Euclidean space and finding the cloest word label in this space\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eJaderberg et al. [11]\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eECCV\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2014\" sdnum=\"2052;\"\u003e2014\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e1) enabling efficient feature sharing for text detection and classification 2) making technical changes over the traditional CNN architectures 3) proposing a method of automated data mining of Flickr.\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eSu and Lu [12]\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eACCV\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2014\" sdnum=\"2052;\"\u003e2014\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eHOG + BLSTM + CTC\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eGordo[13] : Mid-features\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eCVPR\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2015\" sdnum=\"2052;\"\u003e2015\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eproposing local mid-level features for building word image representations\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eJaderberg et al. [14]\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eIJCV\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2015\" sdnum=\"2052;\"\u003e2015\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e1) treating each word as a category and training very large convolutional neural networks to perform word recognition on the whole proposal region 2) generating 9 million images with equal numbers of word samples from a 90k word dictionary\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eJaderberg et al. [15]\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eICLR\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2015\" sdnum=\"2052;\"\u003e2015\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eCNN + CRF\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eShi, Bai, and Yao [16] : CRNN\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eTPAMI\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2017\" sdnum=\"2052;\"\u003e2017\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eCNN + BLSTM + CTC\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eShi et al. [17] : RARE\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eCVPR\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2016\" sdnum=\"2052;\"\u003e2016\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eSTN + CNN + attentional BLSTM\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eLee and Osindero [18] : R2AM\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eCVPR\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2016\" sdnum=\"2052;\"\u003e2016\u003c/td\u003e\n    \u003ctd align=\"center\"\u003epresenting recursive recurrent neural networks with attention modeling\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eLiu et al.  [19] : STAR-Net\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eBMVC\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2016\" sdnum=\"2052;\"\u003e2016\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eSTN + ResNet + BLSTM + CTC\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eLiu et al. [78]\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eICPR\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2016\" sdnum=\"2052;\"\u003e2016\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eintegrating the CNN and WFST classification model\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eMishra et al. [77]\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eCVIU\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2016\" sdnum=\"2052;\"\u003e2016\u003c/td\u003e\n    \u003ctd align=\"center\"\u003echaracter detection (HOG/CNN + SVM +Sliding window) + CRF, combining bottom-up cues from character detection and top-down cues from lexicon\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eSu and Lu [76]\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003ePR\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2017\" sdnum=\"2052;\"\u003e2017\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eHOG(different scale) + BLSTM + CTC (ensemble)\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003e*Yang et al. [20]\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eIJCAI\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2017\" sdnum=\"2052;\"\u003e2017\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e1) CNN + 2D attention-based RNN, applying an auxiliary dense character detection task that helps to learn text specific visual patterns 2) developing a large-scale synthetic dataset\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eYin et al. [21]\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eICCV\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2017\" sdnum=\"2052;\"\u003e2017\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eCNN + CTC\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eWang et al.[66] : GRCNN\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eNIPS\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2017\" sdnum=\"2052;\"\u003e2017\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eGated Recurrent Convulution Layer + BLSTM + CTC\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003e*Cheng et al. [22] : FAN\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eICCV\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2017\" sdnum=\"2052;\"\u003e2017\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e1) proposing the concept of attention drift 2)introducing focusing network to focus deviated attention back on the target areas\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eCheng et al. [23] : AON\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eCVPR\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2018\" sdnum=\"2052;\"\u003e2018\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e1) extracting scene text features in four directions 2) CNN + Attentional BLSTM\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eGao et al. [24]\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eNC\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2019\" sdnum=\"2052;\"\u003e2019\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eattentional ResNet + CNN + CTC\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eLiu et al.  [25] : Char-Net\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eAAAI\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2018\" sdnum=\"2052;\"\u003e2018\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eCNN + STN (facilitating the rectification of individual characters) + LSTM\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003e*Liu et al.  [26] : SqueezedText\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eAAAI\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2018\" sdnum=\"2052;\"\u003e2018\u003c/td\u003e\n    \u003ctd align=\"center\"\u003ebinary convolutional encoder-decoder network + Bi-RNN\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eZhan et al.[73]\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eCVPR\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2018\" sdnum=\"2052;\"\u003e2018\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eCRNN, achieving verisimilar scene text image synthesis by combining three novel designs, including semantic coherence, visual attention, and adaptive text appearance\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003e*Bai et al. [27] : EP\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eCVPR\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2018\" sdnum=\"2052;\"\u003e2018\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eproposing edit probability to effectively handle the misalignment between the training text and the output probability distribution sequence\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eFang et al.[74]\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eMultiMedia\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2018\" sdnum=\"2052;\"\u003e2018\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eResNet + [2D Attentional CNN, CNN-based language module]\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eLiu et al.[75] : EnEsCTC\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eNIPS\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2018\" sdnum=\"2052;\"\u003e2018\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eproposing a novel maximum entropy based regularization for CTC (EnCTC) and an entropy-based pruning method (EsCTC) to effectively reduce the space of the feasible set\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eLiu et al. [28]\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eECCV\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2018\" sdnum=\"2052;\"\u003e2018\u003c/td\u003e\n    \u003ctd align=\"center\"\u003edesigning a multi-task network with an encoder-discriminator-generator architecture to guide the feature of the original image toward that of the clean image\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eWang et al.[61] : MAAN\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eICFHR\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2018\" sdnum=\"2052;\"\u003e2018\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eResNet + BLSTM + Memory-Augmented attentional decoder\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eGao et al. [29]\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eICIP\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2018\" sdnum=\"2052;\"\u003e2018\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eattentional DenseNet + BLSTM + CTC\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eShi et al. [30] : ASTER\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eTPAMI\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2018\" sdnum=\"2052;\"\u003e2018\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eTPS + ResNet + bidirectional attention-based BLSTM\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eChen et al. [60] : ASTER + AEG\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eNC\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2019\" sdnum=\"2052;\"\u003e2019\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eTPS + ResNet + bidirectional attention-based BLSTM + AEG\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eLuo et al. [46] : MORAN\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003ePR\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2019\" sdnum=\"2052;\"\u003e2019\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eMulti-object rectification network + CNN + attentional BLSTM\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eLuo et al. [61] : MORAN-v2\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003ePR\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2019\" sdnum=\"2052;\"\u003e2019\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eMulti-object rectification network + ResNet + attentional BLSTM\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eChen et al. [60] : MORAN-v2 + AEG\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eNC\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2019\" sdnum=\"2052;\"\u003e2019\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eMulti-object rectification network + ResNet + attentional BLSTM + AEG\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eXie et al. [47] : CAN\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eACM\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2019\" sdnum=\"2052;\"\u003e2019\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eResNet + CNN + GLU\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003e*Liao et al.[48] : CA-FCN\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eAAAI\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2019\" sdnum=\"2052;\"\u003e2019\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eperforming character classification at each pixel location and needing character-level annotations\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003e*Li et al. [49] : SAR\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eAAAI\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2019\" sdnum=\"2052;\"\u003e2019\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eResNet + 2D attentional LSTM\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"23\" align=\"center\"\u003eZhan el at. [55]: ESIR\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eCVPR\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2019\" sdnum=\"2052;\"\u003e2019\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eIterative rectification Network + ResNet + attentional BLSTM\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eZhang et al. [56]: SSDAN\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eCVPR\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2019\" sdnum=\"2052;\"\u003e2019\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eattentional CNN + GAS + GRU\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eYang et al. [62]: ScRN\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eICCV\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2019\" sdnum=\"2052;\"\u003e2019\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eSymmetry-constrained Rectification Network + ResNet + BLSTM + attentional GRU\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eWang et al. [64]: GCAM\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eICME\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2019\" sdnum=\"2052;\"\u003e2019\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eConvolutional Block Attention Module (CBAM) + ResNet + BLSTM + the proposed Gated Cascade Attention Module (GCAM)\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eJeonghun et al. [65]\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eICCV\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2019\" sdnum=\"2052;\"\u003e2019\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eTPS + ResNet + BLSTM + Attention Mechanism\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eHuang et al. [67] : EPAN\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eNC\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2019\" sdnum=\"2052;\"\u003e2019\u003c/td\u003e\n    \u003ctd align=\"center\"\u003elearning to sample features from the text region of 2D feature maps and innovatively introducing a two-stage attention mechanism\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eGao et al. [68]\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eNC\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2019\" sdnum=\"2052;\"\u003e2019\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eattentional DenseNET + 4-layer CNN + CTC\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eQi et al. [69] : CCL\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eICDAR\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2019\" sdnum=\"2052;\"\u003e2019\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eResNet + [CTC, CCL]\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eWang et al. [70] : ReELFA\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eICDAR\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2019\" sdnum=\"2052;\"\u003e2019\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eVGG + attentional LSTM, utilizing one-hot encoded coordinates to indicate the spatial relationship of pixels and character center masks to help focus attention on the right feature areas\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eZhu et al. [71] : HATN\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eICIP\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2019\" sdnum=\"2052;\"\u003e2019\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eResNet50 + Hierarchical Attention Mechanism (Transformer structure)\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eZhan et al. [72] : SF-GAN\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eCVPR\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2019\" sdnum=\"2052;\"\u003e2019\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eResNet50 + attentional Decoder, synthesising realistic scene text images for training better recognition models\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eLiao et al. [79] : SAM\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eTPAMI\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2019\" sdnum=\"2052;\"\u003e2019\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eSpatial attentional module (SAM)\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eLiao et al. [79] : seg-SAM\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eTPAMI\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2019\" sdnum=\"2052;\"\u003e2019\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eCharacter segmentation module + Spatial attention module (SAM)\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eWang et al. [80] : DAN\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eAAAI\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2020\" sdnum=\"2052;\"\u003e2020\u003c/td\u003e\n    \u003ctd align=\"center\"\u003edecoupling the decoder of the traditional attention mechanism into a convolutional alignment module and a decoupled text decoder\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eWang et al. [82] : TextSR\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003earXiv\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2019\" sdnum=\"2052;\"\u003e2019\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eattempting to solve small texts with super-resolution methods\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eWan et al. [83] : TextScanner\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eAAAI\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2020\" sdnum=\"2052;\"\u003e2020\u003c/td\u003e\n    \u003ctd align=\"center\"\u003ean effective segmentation-based dual-branch framework for scene text recognition\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eHu et al. [84] : GTC\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eAAAI\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2020\" sdnum=\"2052;\"\u003e2020\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eattempting to use GCN to learn the local correlations of feature sequence\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eLuo et al. [85] \u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eIJCV\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2020\" sdnum=\"2052;\"\u003e2020\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eseparating text content from noisy background styles\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003e*Litman et al. [86]\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eCVPR\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2020\" sdnum=\"2052;\"\u003e2020\u003c/td\u003e\n    \u003ctd align=\"center\"\u003etraining a deep BiLSTM encoder, thus improving the encoding of contextual dependencies\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eYu et al. [87]\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eCVPR\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2020\" sdnum=\"2052;\"\u003e2020\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e introducing a global semantic reasoning module (GSRM) to capture global semantic context through multi-way parallel transmission\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eQiao et al. [101] : SEED\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eCVPR\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2020\" sdnum=\"2052;\"\u003e2020\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e proposing a semantics enhanced encoder-decoder framework to robustly recognize low-quality scene texts\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eBleeker et al. [93] : Bi-STET\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eECAI\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2020\" sdnum=\"2052;\"\u003e2020\u003c/td\u003e\n    \u003ctd align=\"center\"\u003ea novel bidirectional STR method with a single decoder for bidirectional text decoding\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003e*Bartz et al. [94] : KISS\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003earXiv\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2020\" sdnum=\"2052;\"\u003e2020\u003c/td\u003e\n    \u003ctd align=\"center\"\u003ea new model for STR that consists of two ResNet based feature extractors, a spatial transformer, and a transformer\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eZhang et al. [95] : SPIN\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003earXiv\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2020\" sdnum=\"2052;\"\u003e2020\u003c/td\u003e\n    \u003ctd align=\"center\"\u003ea new learnable geometric-unrelated module which allows the color manipulation of source data within the network\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eLin et al. [96] : FASDA\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003earXiv\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2020\" sdnum=\"2052;\"\u003e2020\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eimplementing sequence-level domain adaption for STR\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eZhang et al. [98] : AutoSTR\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eECCV\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2020\" sdnum=\"2052;\"\u003e2020\u003c/td\u003e\n    \u003ctd align=\"center\"\u003esearching data-dependent backbones\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eMou et al. [99] : PlugNet\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eECCV\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2020\" sdnum=\"2052;\"\u003e2020\u003c/td\u003e\n    \u003ctd align=\"center\"\u003ecombining the pluggable super-resolution unit to solve the low-quality text recognition from the feature-level\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003e*Yue et al. [100] : RobustScanner\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e×\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e√\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eECCV\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2020\" sdnum=\"2052;\"\u003e2020\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e mitigating the misrecognition problem of the encoderdecoder with attention framework on contextless text images\u003c/td\u003e\n  \u003c/tr\u003e\n\u003c/table\u003e\n\n\u003ca id=\"22-performance-comparison-on-benchmark-datasets\"\u003e\u003c/a\u003e\n### 2.2 Performance Comparison on Benchmark Datasets\n\nIn this section, we compare the performance of the current advanced algorithms on benchmark datasets, including IIIT5K，SVT，IC03，IC13，SVT-P，CUTE80，IC15，COCO-Text, RCTW-17, MWTI, CTW，SCUT-CTW1500, LSVT, ArT and ReCTS-25k.\n\nIt is notable that 1) The '*' indicates the methods that use the extra datasets other than Synth90k and SynthText. 2) The **bold** represents the best recognition results. 3) '^' denotes the best recognition results of using extra datasets. 4) '@' represents the methods under different evaluation that only uses 1811 test images. 5) 'SK', 'ST', 'ExPu', 'ExPr' and 'Un' indicates the methods that use Synth90K, SynthText, Extra Public Data, Extra Private Data and unknown data, respectively. 6) 'D_A' means data augmentation. 7) IC5-S contains only 1811 cropped text instances.\n\n\u003ca id=\"221-performance-comparison-of-recognition-algorithms-on-regular-latin-datasets\"\u003e\u003c/a\u003e\n#### 2.2.1 Performance Comparison of Recognition Algorithms on Regular Latin Datasets\n\n\u003ctable cellspacing=\"0\" border=\"0\"\u003e\n  \u003ccolgroup width=\"271\"\u003e\u003c/colgroup\u003e\n  \u003ccolgroup span=\"10\" width=\"89\"\u003e\u003c/colgroup\u003e\n  \u003ccolgroup width=\"172\"\u003e\u003c/colgroup\u003e\n  \u003ccolgroup span=\"2\" width=\"86\"\u003e\u003c/colgroup\u003e\n  \u003ctr\u003e\n    \u003ctd colspan=14 height=\"34\" align=\"center\"\u003e\u003cb\u003ePerformance Comparison of Recognition Algorithms on Regular Latin Datasets\u003c/b\u003e\u003c/td\u003e\n    \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd rowspan=2 height=\"39\" align=\"center\"\u003e\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;Method\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u003c/td\u003e\n    \u003ctd colspan=3 align=\"center\"\u003eIIIT5K\u003c/td\u003e\n    \u003ctd colspan=2 align=\"center\"\u003eSVT\u003c/td\u003e\n    \u003ctd colspan=4 align=\"center\"\u003eIC03\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eIC13\u003c/td\u003e\n    \u003ctd rowspan=2 align=\"center\"\u003e\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;Data\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u003c/td\u003e\n    \u003ctd rowspan=2 align=\"center\"\u003eSource\u003c/td\u003e\n    \u003ctd rowspan=2 align=\"center\"\u003eTime\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd align=\"center\" sdval=\"50\" sdnum=\"2052;\"\u003e50\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e1K\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eNone\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"50\" sdnum=\"2052;\"\u003e50\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eNone\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"50\" sdnum=\"2052;\"\u003e50\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eFull\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e50k\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eNone\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eNone\u003c/td\u003e\n    \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eWang et al. [1] : ABBYY\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"24.3\" sdnum=\"2052;\"\u003e24.3\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"35\" sdnum=\"2052;\"\u003e35\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"56\" sdnum=\"2052;\"\u003e56\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"55\" sdnum=\"2052;\"\u003e55\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eUn\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eICCV\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2011\" sdnum=\"2052;\"\u003e2011\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eWang et al. [1] : SYNTH+PLEX\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"57\" sdnum=\"2052;\"\u003e57\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"76\" sdnum=\"2052;\"\u003e76\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"62\" sdnum=\"2052;\"\u003e62\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eExPr\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eICCV\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2011\" sdnum=\"2052;\"\u003e2011\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eMishra et al. [2]\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"64.1\" sdnum=\"2052;\"\u003e64.1\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"57.5\" sdnum=\"2052;\"\u003e57.5\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"73.2\" sdnum=\"2052;\"\u003e73.2\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"81.8\" sdnum=\"2052;\"\u003e81.8\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"67.8\" sdnum=\"2052;\"\u003e67.8\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eExPu\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eBMVC\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2012\" sdnum=\"2052;\"\u003e2012\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eWang et al. [3]\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"70\" sdnum=\"2052;\"\u003e70\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"90\" sdnum=\"2052;\"\u003e90\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"84\" sdnum=\"2052;\"\u003e84\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eExPr\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eICPR\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2012\" sdnum=\"2052;\"\u003e2012\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eGoel et al. [4] : wDTW\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"77.3\" sdnum=\"2052;\"\u003e77.3\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"89.7\" sdnum=\"2052;\"\u003e89.7\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eUn\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eICDAR\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2013\" sdnum=\"2052;\"\u003e2013\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eBissacco et al. [5] : PhotoOCR\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"90.4\" sdnum=\"2052;\"\u003e90.4\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"78\" sdnum=\"2052;\"\u003e78\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"87.6\" sdnum=\"2052;\"\u003e87.6\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eExPr\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eICCV\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2013\" sdnum=\"2052;\"\u003e2013\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003ePhan et al. [6]\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"73.7\" sdnum=\"2052;\"\u003e73.7\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"82.2\" sdnum=\"2052;\"\u003e82.2\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eExPu\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eICCV\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2013\" sdnum=\"2052;\"\u003e2013\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eAlsharif et al. [7] : HMM/Maxout\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"74.3\" sdnum=\"2052;\"\u003e74.3\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"93.1\" sdnum=\"2052;\"\u003e93.1\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"88.6\" sdnum=\"2052;\"\u003e88.6\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"85.1\" sdnum=\"2052;\"\u003e85.1\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eExPu\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eICLR\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2014\" sdnum=\"2052;\"\u003e2014\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eAlmazan et al [8] : KCSR\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"88.6\" sdnum=\"2052;\"\u003e88.6\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"75.6\" sdnum=\"2052;\"\u003e75.6\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"87\" sdnum=\"2052;\"\u003e87\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eExPu\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eTPAMI\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2014\" sdnum=\"2052;\"\u003e2014\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eYao et al. [9] : Strokelets\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"80.2\" sdnum=\"2052;\"\u003e80.2\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"69.3\" sdnum=\"2052;\"\u003e69.3\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"75.9\" sdnum=\"2052;\"\u003e75.9\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"88.5\" sdnum=\"2052;\"\u003e88.5\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"80.3\" sdnum=\"2052;\"\u003e80.3\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eExPu\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eCVPR\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2014\" sdnum=\"2052;\"\u003e2014\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eR.-Serrano et al.[10] : Label embedding\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"76.1\" sdnum=\"2052;\"\u003e76.1\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"57.4\" sdnum=\"2052;\"\u003e57.4\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"70\" sdnum=\"2052;\"\u003e70\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eExPu\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eIJCV\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2015\" sdnum=\"2052;\"\u003e2015\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eJaderberg et al. [11]\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"86.1\" sdnum=\"2052;\"\u003e86.1\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"96.2\" sdnum=\"2052;\"\u003e96.2\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"91.5\" sdnum=\"2052;\"\u003e91.5\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eExPu\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eECCV\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2014\" sdnum=\"2052;\"\u003e2014\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eSu and Lu [12]\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"83\" sdnum=\"2052;\"\u003e83\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"92\" sdnum=\"2052;\"\u003e92\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"82\" sdnum=\"2052;\"\u003e82\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eExPu\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eACCV\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2014\" sdnum=\"2052;\"\u003e2014\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eGordo[13] : Mid-features\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"93.3\" sdnum=\"2052;\"\u003e93.3\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"86.6\" sdnum=\"2052;\"\u003e86.6\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"91.8\" sdnum=\"2052;\"\u003e91.8\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eExPu\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eCVPR\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2015\" sdnum=\"2052;\"\u003e2015\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eJaderberg et al. [14]\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"97.1\" sdnum=\"2052;\"\u003e97.1\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"92.7\" sdnum=\"2052;\"\u003e92.7\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"95.4\" sdnum=\"2052;\"\u003e95.4\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"80.7\" sdnum=\"2052;\"\u003e80.7\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"98.7\" sdnum=\"2052;\"\u003e98.7\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"98.6\" sdnum=\"2052;\"\u003e98.6\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"93.3\" sdnum=\"2052;\"\u003e93.3\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"93.1\" sdnum=\"2052;\"\u003e93.1\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"90.8\" sdnum=\"2052;\"\u003e90.8\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eExPr\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eIJCV\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2015\" sdnum=\"2052;\"\u003e2015\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eJaderberg et al. [15]\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"95.5\" sdnum=\"2052;\"\u003e95.5\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"89.6\" sdnum=\"2052;\"\u003e89.6\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"93.2\" sdnum=\"2052;\"\u003e93.2\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"71.7\" sdnum=\"2052;\"\u003e71.7\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"97.8\" sdnum=\"2052;\"\u003e97.8\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"97\" sdnum=\"2052;\"\u003e97\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"93.4\" sdnum=\"2052;\"\u003e93.4\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"89.6\" sdnum=\"2052;\"\u003e89.6\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"81.8\" sdnum=\"2052;\"\u003e81.8\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eSK + ExPr\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eICLR\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2015\" sdnum=\"2052;\"\u003e2015\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eShi, Bai, and Yao [16] : CRNN\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"97.8\" sdnum=\"2052;\"\u003e97.8\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"95\" sdnum=\"2052;\"\u003e95\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"81.2\" sdnum=\"2052;\"\u003e81.2\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"97.5\" sdnum=\"2052;\"\u003e97.5\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"82.7\" sdnum=\"2052;\"\u003e82.7\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"98.7\" sdnum=\"2052;\"\u003e98.7\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"98\" sdnum=\"2052;\"\u003e98\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"95.7\" sdnum=\"2052;\"\u003e\u003cb\u003e95.7\u003c/b\u003e\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"91.9\" sdnum=\"2052;\"\u003e91.9\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"89.6\" sdnum=\"2052;\"\u003e89.6\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eSK\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eTPAMI\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2017\" sdnum=\"2052;\"\u003e2017\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eShi et al. [17] : RARE\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"96.2\" sdnum=\"2052;\"\u003e96.2\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"93.8\" sdnum=\"2052;\"\u003e93.8\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"81.9\" sdnum=\"2052;\"\u003e81.9\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"95.5\" sdnum=\"2052;\"\u003e95.5\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"81.9\" sdnum=\"2052;\"\u003e81.9\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"98.3\" sdnum=\"2052;\"\u003e98.3\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"96.2\" sdnum=\"2052;\"\u003e96.2\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"94.8\" sdnum=\"2052;\"\u003e94.8\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"90.1\" sdnum=\"2052;\"\u003e90.1\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"88.6\" sdnum=\"2052;\"\u003e88.6\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eSK\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eCVPR\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2016\" sdnum=\"2052;\"\u003e2016\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eLee and Osindero [18] : R2AM\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"96.8\" sdnum=\"2052;\"\u003e96.8\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"94.4\" sdnum=\"2052;\"\u003e94.4\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"78.4\" sdnum=\"2052;\"\u003e78.4\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"96.3\" sdnum=\"2052;\"\u003e96.3\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"80.7\" sdnum=\"2052;\"\u003e80.7\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"97.9\" sdnum=\"2052;\"\u003e97.9\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"97\" sdnum=\"2052;\"\u003e97\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"88.7\" sdnum=\"2052;\"\u003e88.7\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"90\" sdnum=\"2052;\"\u003e90\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eSK\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eCVPR\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2016\" sdnum=\"2052;\"\u003e2016\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eLiu et al.  [19] : STAR-Net\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"97.7\" sdnum=\"2052;\"\u003e97.7\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"94.5\" sdnum=\"2052;\"\u003e94.5\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"83.3\" sdnum=\"2052;\"\u003e83.3\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"95.5\" sdnum=\"2052;\"\u003e95.5\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"83.6\" sdnum=\"2052;\"\u003e83.6\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"96.9\" sdnum=\"2052;\"\u003e96.9\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"95.3\" sdnum=\"2052;\"\u003e95.3\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"89.9\" sdnum=\"2052;\"\u003e89.9\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"89.1\" sdnum=\"2052;\"\u003e89.1\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eSK + ExPr\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eBMVC\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2016\" sdnum=\"2052;\"\u003e2016\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003e*Liu et al. [78]\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"94.1\" sdnum=\"2052;\"\u003e94.1\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"84.7\" sdnum=\"2052;\"\u003e84.7\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"92.5\" sdnum=\"2052;\"\u003e92.5\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"96.8\" sdnum=\"2052;\"\u003e96.8\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"92.2\" sdnum=\"2052;\"\u003e92.2\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eExPu (D_A)\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eICPR\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2016\" sdnum=\"2052;\"\u003e2016\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003e*Mishra et al. [77]\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"78.07\" sdnum=\"2052;\"\u003e78.07\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"46.73\" sdnum=\"2052;\"\u003e46.73\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"78.2\" sdnum=\"2052;\"\u003e78.2\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"88\" sdnum=\"2052;\"\u003e88\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"67.7\" sdnum=\"2052;\"\u003e67.7\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"60.18\" sdnum=\"2052;\"\u003e60.18\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eExPu (D_A)\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eCVIU\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2016\" sdnum=\"2052;\"\u003e2016\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003e*Su and Lu [76]\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"91\" sdnum=\"2052;\"\u003e91\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"95\" sdnum=\"2052;\"\u003e95\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"89\" sdnum=\"2052;\"\u003e89\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"76\" sdnum=\"2052;\"\u003e76\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eSK + ExPu\u003c/td\u003e\n    \u003ctd align=\"center\"\u003ePR\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2017\" sdnum=\"2052;\"\u003e2017\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003e*Yang et al. [20]\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"97.8\" sdnum=\"2052;\"\u003e97.8\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"96.1\" sdnum=\"2052;\"\u003e96.1\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"95.2\" sdnum=\"2052;\"\u003e95.2\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"97.7\" sdnum=\"2052;\"\u003e97.7\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eExPu\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eIJCAI\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2017\" sdnum=\"2052;\"\u003e2017\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eYin et al. [21]\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"98.7\" sdnum=\"2052;\"\u003e98.7\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"96.1\" sdnum=\"2052;\"\u003e96.1\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"78.2\" sdnum=\"2052;\"\u003e78.2\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"95.1\" sdnum=\"2052;\"\u003e95.1\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"72.5\" sdnum=\"2052;\"\u003e72.5\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"97.6\" sdnum=\"2052;\"\u003e97.6\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"96.5\" sdnum=\"2052;\"\u003e96.5\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"81.1\" sdnum=\"2052;\"\u003e81.1\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"81.4\" sdnum=\"2052;\"\u003e81.4\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eSK\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eICCV\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2017\" sdnum=\"2052;\"\u003e2017\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eWang et al.[66] : GRCNN\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"98\" sdnum=\"2052;\"\u003e98\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"95.6\" sdnum=\"2052;\"\u003e95.6\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"80.8\" sdnum=\"2052;\"\u003e80.8\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"96.3\" sdnum=\"2052;\"\u003e96.3\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"81.5\" sdnum=\"2052;\"\u003e81.5\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"98.8\" sdnum=\"2052;\"\u003e98.8\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"97.8\" sdnum=\"2052;\"\u003e97.8\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"91.2\" sdnum=\"2052;\"\u003e91.2\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eSK\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eNIPS\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2017\" sdnum=\"2052;\"\u003e2017\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003e*Cheng et al. [22] : FAN\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"99.3\" sdnum=\"2052;\"\u003e99.3\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"97.5\" sdnum=\"2052;\"\u003e97.5\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"87.4\" sdnum=\"2052;\"\u003e87.4\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"97.1\" sdnum=\"2052;\"\u003e97.1\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"85.9\" sdnum=\"2052;\"\u003e85.9\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"99.2\" sdnum=\"2052;\"\u003e99.2\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"97.3\" sdnum=\"2052;\"\u003e97.3\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"94.2\" sdnum=\"2052;\"\u003e94.2\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"93.3\" sdnum=\"2052;\"\u003e93.3\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eSK + ST (Pixel_wise)\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eICCV\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2017\" sdnum=\"2052;\"\u003e2017\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eCheng et al. [23] : AON\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"99.6\" sdnum=\"2052;\"\u003e\u003cb\u003e99.6\u003c/b\u003e\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"98.1\" sdnum=\"2052;\"\u003e98.1\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"87\" sdnum=\"2052;\"\u003e87\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"96\" sdnum=\"2052;\"\u003e96\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"82.8\" sdnum=\"2052;\"\u003e82.8\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"98.5\" sdnum=\"2052;\"\u003e98.5\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"97.1\" sdnum=\"2052;\"\u003e97.1\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"91.5\" sdnum=\"2052;\"\u003e91.5\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eSK + ST (D_A)\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eCVPR\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2018\" sdnum=\"2052;\"\u003e2018\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eGao et al. [24]\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"99.1\" sdnum=\"2052;\"\u003e99.1\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"97.9\" sdnum=\"2052;\"\u003e97.9\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"81.8\" sdnum=\"2052;\"\u003e81.8\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"97.4\" sdnum=\"2052;\"\u003e97.4\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"82.7\" sdnum=\"2052;\"\u003e82.7\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"98.7\" sdnum=\"2052;\"\u003e98.7\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"96.7\" sdnum=\"2052;\"\u003e96.7\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"89.2\" sdnum=\"2052;\"\u003e89.2\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"88\" sdnum=\"2052;\"\u003e88\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eSK\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eNC\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2019\" sdnum=\"2052;\"\u003e2019\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eLiu et al.  [25] : Char-Net\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"83.6\" sdnum=\"2052;\"\u003e83.6\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"84.4\" sdnum=\"2052;\"\u003e84.4\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"93.3\" sdnum=\"2052;\"\u003e93.3\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"91.5\" sdnum=\"2052;\"\u003e91.5\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"90.8\" sdnum=\"2052;\"\u003e90.8\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eSK (D_A)\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eAAAI\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2018\" sdnum=\"2052;\"\u003e2018\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003e*Liu et al.  [26] : SqueezedText\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"97\" sdnum=\"2052;\"\u003e97\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"94.1\" sdnum=\"2052;\"\u003e94.1\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"87\" sdnum=\"2052;\"\u003e87\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"95.2\" sdnum=\"2052;\"\u003e95.2\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"98.8\" sdnum=\"2052;\"\u003e98.8\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"97.9\" sdnum=\"2052;\"\u003e97.9\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"93.8\" sdnum=\"2052;\"\u003e93.8\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"93.1\" sdnum=\"2052;\"\u003e93.1\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"92.9\" sdnum=\"2052;\"\u003e92.9\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eExPr\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eAAAI\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2018\" sdnum=\"2052;\"\u003e2018\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003e*Zhan et al.[73]\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"98.1\" sdnum=\"2052;\"\u003e98.1\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"95.3\" sdnum=\"2052;\"\u003e95.3\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"79.3\" sdnum=\"2052;\"\u003e79.3\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"96.7\" sdnum=\"2052;\"\u003e96.7\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"81.5\" sdnum=\"2052;\"\u003e81.5\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"87.1\" sdnum=\"2052;\"\u003e87.1\u003c/td\u003e\n    \u003ctd align=\"center\"\u003ePr(5 million)\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eCVPR\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2018\" sdnum=\"2052;\"\u003e2018\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003e*Bai et al. [27] : EP\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"99.5\" sdnum=\"2052;\"\u003e99.5\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"97.9\" sdnum=\"2052;\"\u003e97.9\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"88.3\" sdnum=\"2052;\"\u003e88.3\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"96.6\" sdnum=\"2052;\"\u003e96.6\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"87.5\" sdnum=\"2052;\"\u003e87.5\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"98.7\" sdnum=\"2052;\"\u003e98.7\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"97.9\" sdnum=\"2052;\"\u003e97.9\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"94.6\" sdnum=\"2052;\"\u003e94.6\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"94.4\" sdnum=\"2052;\"\u003e94.4\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eSK + ST (Pixel_wise)\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eCVPR\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2018\" sdnum=\"2052;\"\u003e2018\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eFang et al.[74]\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"98.5\" sdnum=\"2052;\"\u003e98.5\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"96.8\" sdnum=\"2052;\"\u003e96.8\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"86.7\" sdnum=\"2052;\"\u003e86.7\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"97.8\" sdnum=\"2052;\"\u003e97.8\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"86.7\" sdnum=\"2052;\"\u003e86.7\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"99.3\" sdnum=\"2052;\"\u003e\u003cb\u003e99.3\u003c/b\u003e\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"98.4\" sdnum=\"2052;\"\u003e98.4\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"94.8\" sdnum=\"2052;\"\u003e94.8\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"93.5\" sdnum=\"2052;\"\u003e93.5\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eSK + ST\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eMultiMedia\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2018\" sdnum=\"2052;\"\u003e2018\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eLiu et al.[75] : EnEsCTC\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"82\" sdnum=\"2052;\"\u003e82\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"80.6\" sdnum=\"2052;\"\u003e80.6\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"92\" sdnum=\"2052;\"\u003e92\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"90.6\" sdnum=\"2052;\"\u003e90.6\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eSK\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eNIPS\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2018\" sdnum=\"2052;\"\u003e2018\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eLiu et al. [28]\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"97.3\" sdnum=\"2052;\"\u003e97.3\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"96.1\" sdnum=\"2052;\"\u003e96.1\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"89.4\" sdnum=\"2052;\"\u003e89.4\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"96.8\" sdnum=\"2052;\"\u003e96.8\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"87.1\" sdnum=\"2052;\"\u003e87.1\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"98.1\" sdnum=\"2052;\"\u003e98.1\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"97.5\" sdnum=\"2052;\"\u003e97.5\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"94.7\" sdnum=\"2052;\"\u003e94.7\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"94\" sdnum=\"2052;\"\u003e94\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eSK\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eECCV\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2018\" sdnum=\"2052;\"\u003e2018\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eWang et al.[61] : MAAN\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"98.3\" sdnum=\"2052;\"\u003e98.3\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"96.4\" sdnum=\"2052;\"\u003e96.4\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"84.1\" sdnum=\"2052;\"\u003e84.1\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"96.4\" sdnum=\"2052;\"\u003e96.4\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"83.5\" sdnum=\"2052;\"\u003e83.5\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"97.4\" sdnum=\"2052;\"\u003e97.4\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"96.4\" sdnum=\"2052;\"\u003e96.4\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"92.2\" sdnum=\"2052;\"\u003e92.2\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"91.1\" sdnum=\"2052;\"\u003e91.1\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eSK\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eICFHR\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2018\" sdnum=\"2052;\"\u003e2018\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eGao et al. [29]\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"99.1\" sdnum=\"2052;\"\u003e99.1\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"97.2\" sdnum=\"2052;\"\u003e97.2\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"83.6\" sdnum=\"2052;\"\u003e83.6\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"97.7\" sdnum=\"2052;\"\u003e97.7\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"83.9\" sdnum=\"2052;\"\u003e83.9\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"98.6\" sdnum=\"2052;\"\u003e98.6\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"96.6\" sdnum=\"2052;\"\u003e96.6\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"91.4\" sdnum=\"2052;\"\u003e91.4\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"89.5\" sdnum=\"2052;\"\u003e89.5\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eSK\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eICIP\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2018\" sdnum=\"2052;\"\u003e2018\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eShi et al. [30] : ASTER\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"99.6\" sdnum=\"2052;\"\u003e\u003cb\u003e99.6\u003c/b\u003e\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"98.8\" sdnum=\"2052;\"\u003e98.8\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"93.4\" sdnum=\"2052;\"\u003e93.4\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"97.4\" sdnum=\"2052;\"\u003e97.4\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"89.5\" sdnum=\"2052;\"\u003e89.5\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"98.8\" sdnum=\"2052;\"\u003e98.8\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"98\" sdnum=\"2052;\"\u003e98\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"94.5\" sdnum=\"2052;\"\u003e94.5\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"91.8\" sdnum=\"2052;\"\u003e91.8\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eSK + ST\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eTPAMI\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2018\" sdnum=\"2052;\"\u003e2018\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eChen et al. [60] : ASTER + AEG\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"99.5\" sdnum=\"2052;\"\u003e99.5\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"98.5\" sdnum=\"2052;\"\u003e98.5\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"94.4\" sdnum=\"2052;\"\u003e94.4\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"97.4\" sdnum=\"2052;\"\u003e97.4\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"90.3\" sdnum=\"2052;\"\u003e90.3\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"99\" sdnum=\"2052;\"\u003e99\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"98.3\" sdnum=\"2052;\"\u003e98.3\u003c/td\u003e\n    \u003ctd align=\"center\"\u003e-\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"95.2\" sdnum=\"2052;\"\u003e95.2\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"95\" sdnum=\"2052;\"\u003e95\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eSK + ST\u003c/td\u003e\n    \u003ctd align=\"center\"\u003eNC\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"2019\" sdnum=\"2052;\"\u003e2019\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd height=\"20\" align=\"center\"\u003eLuo et al. [46] : MORAN\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"97.9\" sdnum=\"2052;\"\u003e97.9\u003c/td\u003e\n    \u003ctd align=\"center\" sdval=\"96.2\" sdnum=\"2052;\"\u003e96.2\u003c/td\u003e\n    \u003ctd align=\"","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2FHCIILAB%2FScene-Text-Recognition","html_url":"https://awesome.ecosyste.ms/projects/github.com%2FHCIILAB%2FScene-Text-Recognition","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2FHCIILAB%2FScene-Text-Recognition/lists"}