{"id":13528867,"url":"https://github.com/ethanhe42/epipolar-transformers","last_synced_at":"2025-04-05T12:07:08.742Z","repository":{"id":41321835,"uuid":"251161090","full_name":"ethanhe42/epipolar-transformers","owner":"ethanhe42","description":"Epipolar Transformers (best paper award, CVPR 2020 workshop)","archived":false,"fork":false,"pushed_at":"2024-05-02T06:33:41.000Z","size":6800,"stargazers_count":412,"open_issues_count":4,"forks_count":37,"subscribers_count":22,"default_branch":"master","last_synced_at":"2024-08-06T10:12:30.561Z","etag":null,"topics":["3d","3dposeestimation","deep-learning","pose-estimation","pytorch"],"latest_commit_sha":null,"homepage":"https://yihui.dev/epipolar-transformers","language":"Jupyter Notebook","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"mit","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/ethanhe42.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":".github/FUNDING.yml","license":"LICENSE","code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null},"funding":{"github":null,"patreon":null,"open_collective":null,"ko_fi":"yihuihe","tidelift":null,"community_bridge":null,"liberapay":null,"issuehunt":null,"otechie":null,"lfx_crowdfunding":null,"custom":null}},"created_at":"2020-03-30T00:09:57.000Z","updated_at":"2024-08-06T10:12:46.911Z","dependencies_parsed_at":"2024-08-06T10:12:34.445Z","dependency_job_id":null,"html_url":"https://github.com/ethanhe42/epipolar-transformers","commit_stats":{"total_commits":25,"total_committers":3,"mean_commits":8.333333333333334,"dds":0.24,"last_synced_commit":"21aa1f3ecfba02489272e3d024d00a2955f1ab6c"},"previous_names":["ethanhe42/epipolar-transformers","yihui-he/epipolar-transformers"],"tags_count":1,"template":false,"template_full_name":null,"repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/ethanhe42%2Fepipolar-transformers","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/ethanhe42%2Fepipolar-transformers/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/ethanhe42%2Fepipolar-transformers/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/ethanhe42%2Fepipolar-transformers/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/ethanhe42","download_url":"https://codeload.github.com/ethanhe42/epipolar-transformers/tar.gz/refs/heads/master","host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":247332609,"owners_count":20921853,"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":["3d","3dposeestimation","deep-learning","pose-estimation","pytorch"],"created_at":"2024-08-01T07:00:26.747Z","updated_at":"2025-04-05T12:07:08.716Z","avatar_url":"https://github.com/ethanhe42.png","language":"Jupyter Notebook","funding_links":["https://ko-fi.com/yihuihe"],"categories":["5. Learning based SLAM"],"sub_categories":["5.2 Others"],"readme":"# Epipolar Transformers\n\n![Screen_Shot_2022-10-17_at_5 46 06_PM](https://github.com/ethanhe42/epipolar-transformers/assets/10027339/591a9fe5-bd8e-446b-ae91-0a2d3c0eb2b4)\n\n\n[GitHub - yihui-he/epipolar-transformers: Epipolar Transformers (CVPR 2020)](https://github.com/yihui-he/epipolar-transformers)\n\n[Epipolar Transformers](https://arxiv.org/abs/2005.04551)\n\n[Yihui He](http://yihui-he.github.io/), [Rui Yan](https://github.com/Yre), [Katerina Fragkiadaki](https://www.cs.cmu.edu/~katef/), [Shoou-I Yu](https://sites.google.com/view/shoou-i-yu/home) (Carnegie Mellon University, Facebook Reality Labs)\n\n**CVPR 2020**, **[CVPR workshop Best Paper Award](https://vuhcs.github.io/)**\n\nOral presentation and human pose demo videos ([playlist](https://www.youtube.com/playlist?list=PLkz610aVEqV-f4Ws0pH0e8Nm_2wTGI1yP)):\n\n[https://www.youtube.com/embed/nfb0kfVWjcs](https://www.youtube.com/embed/nfb0kfVWjcs)\n\n[https://www.youtube.com/embed/ig5c-qTaYkg](https://www.youtube.com/embed/ig5c-qTaYkg)\n\n## Models\n\n| config | MPJPE (mm) | model \u0026 log |\n| --- | --- | --- |\n| https://www.notion.soconfigs/benchmark/keypoint_h36m.yaml | 45.3 | https://github.com/yihui-he/epipolar-transformers/releases/download/outputs/outs.benchmark.keypoint_h36m_afterfix.zip |\n| https://www.notion.soconfigs/epipolar/keypoint_h36m_zresidual_fixed.yaml | 33.1 | https://github.com/yihui-he/epipolar-transformers/releases/download/outputs/outs.epipolar.keypoint_h36m_fixed.zip |\n| https://www.notion.soconfigs/epipolar/keypoint_h36m_zresidual_aug.yaml | 30.4 | https://github.com/yihui-he/epipolar-transformers/releases/download/outputs/outs.epipolar.keypoint_h36m_fixed_aug.zip |\n| https://www.notion.soconfigs/epipolar/keypoint_h36m_resnet152_384_pretrained_8gpu.yaml | 19 |  |\n\nWe also provide 2D to 3D lifting network implementations for these two papers: \n\n- [3D Hand Shape and Pose from Images in the Wild](https://arxiv.org/abs/1902.03451), CVPR 2019\n    - `configs/lifting/img_lifting_rot_h36m.yaml` (Human 3.6M)\n    - `configs/lifting/img_lifting_rot.yaml` (RHD)\n- [Learning to Estimate 3D Hand Pose from Single RGB Images](http://openaccess.thecvf.com/content_ICCV_2017/papers/Zimmermann_Learning_to_Estimate_ICCV_2017_paper.pdf), ICCV 2017\n    - `configs/lifting/lifting_direct_h36m.yaml` (Human 3.6M)\n    - `configs/lifting/lifting_direct.yaml` (RHD)\n\n## Setup\n\n### Requirements\n\nPython 3, pytorch \u003e 1.2+ and pytorch \u003c 1.4\n\n```\npip install -r requirements.txtconda install pytorch cudatoolkit=10.0 -c pytorch\n```\n\n### Pretrained weights download\n\n```\nmkdir outscd datasets/bash get_pretrained_models.sh\n```\n\nPlease follow the instructions in `datasets/README.md` for preparing the dataset\n\n### Training\n\n```\npython main.py --cfg path/to/configtensorboard --logdir outs/\n```\n\n### Testing\n\nTesting with latest checkpoints\n\n```\npython main.py --cfg configs/xxx.yaml DOTRAIN False\n```\n\nTesting with weights\n\n```\npython main.py --cfg configs/xxx.yaml DOTRAIN False WEIGHTS xxx.pth\n```\n\n## Visualization\n\n### Epipolar Transformers Visualization\n\n![https://raw.githubusercontent.com/yihui-he/epipolar-transformers/master/assets/et_vis.png](https://raw.githubusercontent.com/yihui-he/epipolar-transformers/master/assets/et_vis.png)\n\n- Download the [output pkls](https://github.com/yihui-he/epipolar-transformers/releases/download/outputs/outs.epipolar.keypoint_h36m_fixed.visualizations.zip) for non-augmented models and extract under `outs/`\n- Make sure `outs/epipolar/keypoint_h36m_fixed/visualizations/h36m/output_1.pkl` exists.\n- Use `[scripts/vis_hm36_score.ipynb](https://github.com/yihui-he/epipolar-transformers/blob/master/scripts/vis_hm36_score.ipynb)`\n    - To select a point, click on the reference view (upper left), the source view along with corresponding epipolar line, and the peaks for different feature matchings are shown at the bottom left.\n\n### Human 3.6M input visualization\n\n![https://raw.githubusercontent.com/yihui-he/epipolar-transformers/master/assets/h36m_vis.png](https://raw.githubusercontent.com/yihui-he/epipolar-transformers/master/assets/h36m_vis.png)\n\n```bash\npython main.py --cfg configs/epipolar/keypoint_h36m.yaml DOTRAIN False DOTEST False EPIPOLAR.VIS True  VIS.H36M True SOLVER.IMS_PER_BATCH 1\npython main.py --cfg configs/epipolar/keypoint_h36m.yaml DOTRAIN False DOTEST False VIS.MULTIVIEWH36M True EPIPOLAR.VIS True SOLVER.IMS_PER_BATCH 1\n```\n\n### Human 3.6M prediction visualization\n\n[https://www.youtube.com/embed/ig5c-qTaYkg](https://www.youtube.com/embed/ig5c-qTaYkg)\n\n```bash\n# generate images\npython main.py --cfg configs/epipolar/keypoint_h36m_zresidual_fixed.yaml DOTRAIN False DOTEST True VIS.VIDEO True DATASETS.H36M.TEST_SAMPLE 2\n# generate images\npython main.py --cfg configs/benchmark/keypoint_h36m.yaml DOTRAIN False DOTEST True VIS.VIDEO True DATASETS.H36M.TEST_SAMPLE 2\n# use https://github.com/yihui-he/multiview-human-pose-estimation-pytorch to generate images for ICCV 19\npython run/pose2d/valid.py --cfg experiments-local/mixed/resnet50/256_fusion.yaml \n# set test batch size to 1 and PRINT_FREQ to 2\n# generate video\npython scripts/video.py --src outs/epipolar/keypoint_h36m_fixed/video/multiview_h36m_val/\n```\n\n## Citing Epipolar Transformers\n\nIf you find Epipolar Transformers helps your research, please cite the paper:\n\n```\n@inproceedings{epipolartransformers,\n  title={Epipolar Transformers},\n  author={He, Yihui and Yan, Rui and Fragkiadaki, Katerina and Yu, Shoou-I},\n  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},\n  pages={7779--7788},\n  year={2020}\n}\n```\n\n## FAQ\n\nPlease create a new issue:\n\n[Issues · yihui-he/epipolar-transformers](https://github.com/yihui-he/epipolar-transformers/issues)\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fethanhe42%2Fepipolar-transformers","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fethanhe42%2Fepipolar-transformers","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fethanhe42%2Fepipolar-transformers/lists"}