Ecosyste.ms: Awesome

An open API service indexing awesome lists of open source software.

Awesome Lists | Featured Topics | Projects

https://github.com/beacandler/EATEN

EATEN: Entity-aware Attention for Single Shot Visual Text Extraction
https://github.com/beacandler/EATEN

Last synced: 2 months ago
JSON representation

EATEN: Entity-aware Attention for Single Shot Visual Text Extraction

Host: GitHub
URL: https://github.com/beacandler/EATEN
Owner: beacandler
Created: 2019-05-03T05:52:28.000Z (over 5 years ago)
Default Branch: master
Last Pushed: 2019-12-29T09:14:26.000Z (about 5 years ago)
Last Synced: 2024-08-03T12:15:45.204Z (5 months ago)
Size: 3.49 MB
Stars: 173
Watchers: 13
Forks: 17
Open Issues: 7
Metadata Files:
- Readme: README.md

Awesome Lists containing this project

awesome-key-information-extraction - [link

README

        # EATEN: Entity-aware Attention for Single Shot Visual Text Extraction     

Accepted to ICDAR 2019 [arxiv](https://arxiv.org/abs/1909.09380#)      

Authors: He Guo, Xiameng Qin, Jiaming Liu, Junyu Han, Jingtuo Liu and Errui Ding     

## Abstract

This repository is designed to provide an open-source dataset for Visual Text Extraction.

## Samples

### Train ticket

#### Real images 

![real1](./figures/real1.jpg)

![real2](./figures/real2.jpg)

#### Synthetic images 

##### Some clean images

![synth-easy](./figures/synth-easy.png)

##### Some hard images

![synth-hard](./figures/synth-hard.png)

### Passport

##### Some images

![passport-easy](./figures/passport-easy.png)

##### Some hard images

![passport-hard](./figures/passport-hard.png)

### Business card

![bc1](./figures/bc1.png)

![bc2](./figures/bc2.png)

## Downloads

The dataset can be downloaded through the following link:   

[baiduyun](https://pan.baidu.com/s/1HVMa_bpCeegticZVFOkJ5g), PASSWORD: e4z1 

Some details:         

|scenes| number | size| Google Drive link |

|-------------------|:-------------------:|:---------------------:|:---------------------:|

|train ticket | 300k synth + 1.9 real| 13G|[dataset_trainticket.tar](https://drive.google.com/open?id=1zxu44zSBtBw9CZjIIVPfK0h9YqHfQKID)|  

|passport | 100k synth |5.8G|[dataset_passport.tar](https://drive.google.com/open?id=11XhKsjqzZY6jBakkLDy8lywE3w37kPaC)|  

|business card | 200k synth| 19G|[dataset_business.tar.0   ](https://drive.google.com/open?id=1irnwv8MqK9nCT_NqmnjHfQ8DlwdMwcBs)[dataset_business.tar.1   ](https://drive.google.com/open?id=16Y26c5VCBYx_CD7Mz3UR6rCbVclq7sRP)[dataset_business.tar.2   ](https://drive.google.com/open?id=1ilQC0cmLhW_N9sl6UbGQZHlk5p_H98Q4)[dataset_business.tar.3   ](https://drive.google.com/open?id=1URl4V8Zxgpl0jifkjIDjy-gMa5g3Jtsj)|   

   

## Limitations&&Todo

  - [A large of training data]            

    Todo: 

    1. Use CycleGan or domain adaptation to synth data to train EATEN.

    2. Introduce datasets of STR to EATEN.

  - [Generalization on complex scenes]         

    Todo:

    1. Add bounding box annotations of ToIs to EATEN, such as [2019-ICCV-oral Towards Unconstrained End-to-End Text Spotting](http://openaccess.thecvf.com/content_ICCV_2019/papers/Qin_Towards_Unconstrained_End-to-End_Text_Spotting_ICCV_2019_paper.pdf).

  - [Engineering]         

    1. Merge server decoder to one.

    2. parallel decoding.