{"id":13521746,"url":"https://github.com/MasterBin-IIAU/AlphaRefine","last_synced_at":"2025-03-31T20:32:47.794Z","repository":{"id":37958976,"uuid":"276833649","full_name":"MasterBin-IIAU/AlphaRefine","owner":"MasterBin-IIAU","description":"Official implementation for the CVPR2021 paper Alpha-Refine","archived":false,"fork":false,"pushed_at":"2023-10-03T21:36:42.000Z","size":13963,"stargazers_count":188,"open_issues_count":20,"forks_count":31,"subscribers_count":10,"default_branch":"master","last_synced_at":"2024-11-02T05:32:59.575Z","etag":null,"topics":["alpha-refine","cvpr2021","object-tracking","refinement-module","vot2020","winner"],"latest_commit_sha":null,"homepage":"","language":"Python","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"gpl-3.0","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/MasterBin-IIAU.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":"LICENSE","code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null}},"created_at":"2020-07-03T07:13:39.000Z","updated_at":"2024-05-15T23:04:47.000Z","dependencies_parsed_at":"2023-01-20T23:45:13.043Z","dependency_job_id":"1d7e9a69-c2a8-4bc6-aa56-f48307cba057","html_url":"https://github.com/MasterBin-IIAU/AlphaRefine","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/MasterBin-IIAU%2FAlphaRefine","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/MasterBin-IIAU%2FAlphaRefine/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/MasterBin-IIAU%2FAlphaRefine/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/MasterBin-IIAU%2FAlphaRefine/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/MasterBin-IIAU","download_url":"https://codeload.github.com/MasterBin-IIAU/AlphaRefine/tar.gz/refs/heads/master","host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":246536254,"owners_count":20793411,"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":["alpha-refine","cvpr2021","object-tracking","refinement-module","vot2020","winner"],"created_at":"2024-08-01T06:00:37.749Z","updated_at":"2025-03-31T20:32:44.123Z","avatar_url":"https://github.com/MasterBin-IIAU.png","language":"Python","funding_links":[],"categories":["Papers"],"sub_categories":["CVPR 2021"],"readme":"# Alpha-Refine\n\nThis is the official implementation of [Alpha-Refine: Boosting Tracking Performance by Precise Bounding Box Estimation\n](https://arxiv.org/abs/2012.06815).\n![Architecture](doc/asset/AR-Architecture.png)\n\n## News\n- :warning: We provide a concise script [demo.py](demo.py) as an example of applying alpha refine to dimp. \n**We recommend taking this script as the starting point of exploring our project**.\n- A TensorRT optimized version of AlphaRefine is available [here](https://github.com/ymzis69/AlphaRefine_TensorRT).\n- The code for **CVPR2021** is updated. The old version is still available by\n        \n        git clone -b vot2020 https://github.com/MasterBin-IIAU/AlphaRefine.git \n        \n- AlphaRefine is accepted by the **CVPR2021**\n- :trophy: **Alpha-Refine wins VOT2020 Real-Time Challenge with EAOMultistart 0.499!** \n- VOT2020 winner presentation [slide](VOT20-RT-Report.pdf) has been uploaded.\n\n\n## Setup Alpha-Refine\n\n* **Install AlphaRefine**\n  \n```bash\ngit clone https://github.com/MasterBin-IIAU/AlphaRefine.git\ncd AlphaRefine\n```\nRun the installation script to install all the dependencies. You need to provide the `${conda_install_path}`\n(e.g. `~/anaconda3`) and the name `${env_name}` for the created conda environment (e.g. `alpha`).\n```\n# install dependencies\nbash install.sh ${conda_install_path} ${env_name}\nconda activate alpha\npython setup.py develop\n```  \n\n* **Download AlphaRefine Models**\n\nWe provide the models of *AlphaRefine* here. The **AUC** and **Latency** are tested with SiamRPN++ as the base tracker\non *LaSOT* dataset, using a RTX 2080Ti GPU.\n\nWe recommend download the model into `ltr/checkpoints/ltr/SEx_beta`. \n\n| Tracker        | Backbone         | Latency     | AUC(%)   |  Model  |\n|:--------------:|:----------------:|:-----------:|:-----------:|:----------------:|\n| AR34\u003csub\u003ec+m\u003c/sub\u003e | ResNet34     |  5.1ms  |  55.9  |   [google](https://drive.google.com/file/d/1drLqNq4r9g4ZqGtOGuuLCmHJDh20Fu1m/view?usp=sharing)/[baidu](https://pan.baidu.com/s/1ZCJKk1mXE_96BEpwGiEuMQ)[key:jl1m]|\n| AR18\u003csub\u003ec+m\u003c/sub\u003e | ResNet18     |  4.2ms  |  55.0  |   [google](https://drive.google.com/file/d/1ANf0KCvlFBbGQPpvT-3WNiy414ANkgLZ/view?usp=sharing)/[baidu](https://pan.baidu.com/s/1IIaRNkFVPSG1s71g255CHw)[key:83ef]|\n\nWhen combined with more powerful base trackers, *AlphaRefine* leads to very competitive tracking systems (e.g. *ARDiMP*). \nFollowing are some of the best performed trackers on LaSOT. Results are present in [Performance](#performance)\n\n* **Demo**\n\nWe provide a concise [demo.py](demo.py) as an example for applying alpha refine to dimp.\n**We recommend you should take this script as the starting point of exploring our project**.\nYou may need  [doc/Reproduce.md](doc/Reproduce.md) for setting up the base trackers of our experiments.\n\n## How to apply Alpha-Refine to Your Own Tracker\nWe provide a concise [demo.py](demo.py) as an example for applying alpha refine to dimp.\n\n\n## How to Train Alpha-Refine\nPlease refer to [doc/TRAIN.md](doc/TRAIN.md) for the guidance of training Alpha-Refine.\n\nAfter training, you can refer to [doc/Reproduce.md](doc/Reproduce.md) for reproducing our experiment result.\n\n## Performance\n\nWhen combined with more powerful base trackers, \n*AlphaRefine* leads to very competitive tracking systems (e.g. *ARDiMP*).\nFor more performance reports, please refer to our [paper](https://arxiv.org/abs/2012.06815).\n**You can refer to [doc/Reproduce.md](doc/Reproduce.md) for reproducing our result.**\n\n* **LaSOT**\n\n     | Tracker                   | Success Score    | Speed (fps) | Paper/Code |\n     |:-----------               |:----------------:|:----------------:|:----------------:|\n     | ARDiMP (ours)             | 0.654  |  32 (RTX 2080Ti)  |   [Paper](https://arxiv.org/abs/2012.06815)/[Result](https://drive.google.com/file/d/1UNPwz7qP8SeBTxHF_Cw0JLmrN1jTqJJE/view?usp=sharing) |\n     | Siam R-CNN (CVPR20)       | 0.648  |  5 (Tesla V100)   |   [Paper](https://arxiv.org/pdf/1911.12836.pdf)/[Code](https://github.com/VisualComputingInstitute/SiamR-CNN) |\n     | DimpSuper                 | 0.631  |  39 (RTX 2080Ti)  |   [Paper](https://arxiv.org/pdf/2003.12565.pdf)/[Code](https://github.com/visionml/pytracking)  |\n     | ARDiMP50 (ours)           | 0.602  |  46 (RTX 2080Ti)  |   [Paper](https://arxiv.org/abs/2012.06815)/[Result](https://drive.google.com/file/d/1wJc_-1lCxeGlqEAKd1qER1x_4bWAhujv/view?usp=sharing)  |\n     | PrDiMP50 (CVPR20)         | 0.598  |  30 (Unkown GPU)  |   [Paper](https://arxiv.org/pdf/2003.12565.pdf)/[Code](https://github.com/visionml/pytracking)  |\n     | LTMU (CVPR20)             | 0.572  |  13 (RTX 2080Ti)  |   [Paper](https://arxiv.org/abs/2004.00305)/[Code](https://github.com/Daikenan/LTMU) |\n     | DiMP50 (ICCV19)           | 0.568  |  59 (RTX 2080Ti)  |   [Paper](https://arxiv.org/pdf/1904.07220.pdf)/[Code](https://github.com/visionml/pytracking)  |\n     | Ocean (ECCV20)            | 0.560  |  25 (Tesla V100)  |   [Paper](https://arxiv.org/abs/2006.10721)/[Code](https://github.com/researchmm/TracKit) |  \n     | ARSiamRPN (ours)          | 0.560  |  50 (RTX 2080Ti)  |   [Paper](https://arxiv.org/abs/2012.06815)/[Result](https://drive.google.com/file/d/1u-ou43O_RU9oRFx1UKjzeYe6e-4qnMZZ/view?usp=sharing) |  \n     | SiamAttn (CVPR20)         | 0.560  |  45 (RTX 2080Ti)  |   [Paper](https://arxiv.org/pdf/2004.06711.pdf)/[Code]() |\n     | SiamFC++GoogLeNet (AAAI20)| 0.544  |  90 (RTX 2080Ti)  |   [Paper](https://arxiv.org/pdf/1911.06188.pdf)/[Code](https://github.com/MegviiDetection/video_analyst) |\n     | MAML-FCOS (CVPR20)        | 0.523  |  42 (NVIDIA P100) |   [Paper](https://arxiv.org/pdf/2004.00830.pdf)/[Code]() |\n     | GlobalTrack (AAAI20)      | 0.521  |  6 (GTX TitanX)   |   [Paper](https://arxiv.org/abs/1912.08531)/[Code](https://github.com/huanglianghua/GlobalTrack) |\n     | ATOM (CVPR19)             | 0.515  |  30 (GTX 1080)    |   [Paper](https://arxiv.org/pdf/1811.07628.pdf)/[Code](https://github.com/visionml/pytracking)  |\n\n\n## Acknowledgments\n* This repo is based on [Pytracking](https://github.com/visionml/pytracking.git) which is an exellent work.\n* Thanks for [pysot](https://github.com/STVIR/pysot) and [RTMDNet](https://github.com/IlchaeJung/RT-MDNet) from which\n we borrow the code as base trackers.\n\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2FMasterBin-IIAU%2FAlphaRefine","html_url":"https://awesome.ecosyste.ms/projects/github.com%2FMasterBin-IIAU%2FAlphaRefine","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2FMasterBin-IIAU%2FAlphaRefine/lists"}