{"id":17260370,"url":"https://github.com/mkocabas/pare","last_synced_at":"2025-04-05T12:08:42.262Z","repository":{"id":40628152,"uuid":"327558695","full_name":"mkocabas/PARE","owner":"mkocabas","description":"Code for ICCV2021 paper PARE: Part Attention Regressor for 3D Human Body Estimation","archived":false,"fork":false,"pushed_at":"2024-04-22T22:35:35.000Z","size":21036,"stargazers_count":392,"open_issues_count":22,"forks_count":75,"subscribers_count":14,"default_branch":"master","last_synced_at":"2025-03-29T11:09:56.551Z","etag":null,"topics":["3d-human-reconstruction","3d-human-shape-and-pose-estimation","computer-graphics","computer-vision","human-pose-estimation"],"latest_commit_sha":null,"homepage":"","language":"Python","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"other","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/mkocabas.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,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null}},"created_at":"2021-01-07T09:08:25.000Z","updated_at":"2025-03-19T10:39:01.000Z","dependencies_parsed_at":"2024-10-29T18:23:31.109Z","dependency_job_id":null,"html_url":"https://github.com/mkocabas/PARE","commit_stats":{"total_commits":6,"total_committers":1,"mean_commits":6.0,"dds":0.0,"last_synced_commit":"5278450e08189dbc25487a28d93c13942182ed6a"},"previous_names":[],"tags_count":1,"template":false,"template_full_name":null,"repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/mkocabas%2FPARE","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/mkocabas%2FPARE/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/mkocabas%2FPARE/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/mkocabas%2FPARE/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/mkocabas","download_url":"https://codeload.github.com/mkocabas/PARE/tar.gz/refs/heads/master","host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":247332612,"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-human-reconstruction","3d-human-shape-and-pose-estimation","computer-graphics","computer-vision","human-pose-estimation"],"created_at":"2024-10-15T07:48:02.044Z","updated_at":"2025-04-05T12:08:42.238Z","avatar_url":"https://github.com/mkocabas.png","language":"Python","funding_links":[],"categories":[],"sub_categories":[],"readme":"# PARE: Part Attention Regressor for 3D Human Body Estimation [ICCV 2021]\n\n[![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)]()\n[![report](https://img.shields.io/badge/Project-Page-blue)](https://pare.is.tue.mpg.de/)\n[![report](https://img.shields.io/badge/ArXiv-Paper-red)](https://arxiv.org/abs/2104.08527)\n\n\u003e [**PARE: Part Attention Regressor for 3D Human Body Estimation**](https://arxiv.org/abs/2104.08527),            \n\u003e [Muhammed Kocabas](https://ps.is.tuebingen.mpg.de/person/mkocabas), \n\u003e [Chun-Hao Paul Huang](https://ps.is.tuebingen.mpg.de/person/chuang2),\n\u003e [Otmar Hilliges](https://ait.ethz.ch/people/hilliges/)\n[Michael J. Black](https://ps.is.tuebingen.mpg.de/person/black),        \n\u003e *International Conference on Computer Vision (ICCV), 2021*\n\n\u003cp float=\"left\"\u003e\n  \u003cimg src=\"docs/assets/vibe_vs_pare_p1.gif\" width=\"49%\" /\u003e\n  \u003cimg src=\"docs/assets/vibe_vs_pare_p2.gif\" width=\"49%\" /\u003e\n\n\u003c/p\u003e\n\n## Features\n\nPARE is an occlusion-robust human pose and shape estimation method. This implementation includes the demo and evaluation code for \nPARE implemented in PyTorch.\n\n## Updates\n\n- 13/10/2021: Demo and evaluation code is released.\n\n## Getting Started\n\nPARE has been implemented and tested on Ubuntu 18.04 with \npython \u003e= 3.7. If you don't have a suitable device, \ntry running our Colab demo.\n\nClone the repo:\n\n```shell\ngit clone https://github.com/mkocabas/PARE.git\n```\n\nInstall the requirements using virtualenv or conda:\n\n```shell\n# pip\nsource scripts/install_pip.sh\n\n# conda\nsource scripts/install_conda.sh\n```\n\n## Demo\n\nFirst, you need to download the required data \n(i.e our trained model and SMPL model parameters). It is approximately 1.3GB. \nTo do this you can just run:\n\n```shell\nsource scripts/prepare_data.sh\n```\n\n### Video Demo\nRun the command below. See `scripts/demo.py` for more options.\n```shell script\npython scripts/demo.py --vid_file data/sample_video.mp4 --output_folder logs/demo \n```\n\nSample demo output:\n\n\u003cp float=\"left\"\u003e\n  \u003cimg src=\"docs/assets/demo_output.gif\" width=\"30%\" /\u003e\n\u003c/p\u003e\n\n### Image Folder Demo\n\n```shell script\npython scripts/demo.py --image_folder \u003cpath to image folder\u003e --output_folder logs/demo\n```\n\n#### Output format\n\nIf demo finishes succesfully, it needs to create a file named `pare_output.pkl` in the `--output_folder`.\nWe can inspect what this file contains by:\n\n```\n\u003e\u003e\u003e import joblib # you may also use native pickle here as well\n\n\u003e\u003e\u003e output = joblib.load('pare_output.pkl') \n\n\u003e\u003e\u003e print(output.keys())  \n                                                                                                                                                                                                                                                                                                                                                                                              \ndict_keys([1, 2, 3, 4]) # these are the track ids for each subject appearing in the video\n\n\u003e\u003e\u003e for k,v in output[1].items(): print(k,v.shape) \n\npred_cam (n_frames, 3)          # weak perspective camera parameters in cropped image space (s,tx,ty)\norig_cam (n_frames, 4)          # weak perspective camera parameters in original image space (sx,sy,tx,ty)\nverts (n_frames, 6890, 3)       # SMPL mesh vertices\npose (n_frames, 72)             # SMPL pose parameters\nbetas (n_frames, 10)            # SMPL body shape parameters\njoints3d (n_frames, 49, 3)      # SMPL 3D joints\njoints2d (n_frames, 21, 3)      # 2D keypoint detections by STAF if pose tracking enabled otherwise None\nbboxes (n_frames, 4)            # bbox detections (cx,cy,w,h)\nframe_ids (n_frames,)           # frame ids in which subject with tracking id #1 appears\nsmpl_joints2d (n_frames, 49, 2) # SMPL 2D joints \n```\n## Google Colab\n\n## Training\n\nTraining instructions will follow soon.\n\n## Evaluation\nYou need to download [3DPW](https://virtualhumans.mpi-inf.mpg.de/3DPW/) \nand [3DOH](https://www.yangangwang.com/papers/ZHANG-OOH-2020-03.html) \ndatasets before running the evaluation script. \nAfter the download, the `data` folder should look like:\n\n```shell\ndata/\n├── body_models\n│   └── smpl\n├── dataset_extras\n├── dataset_folders\n│   ├── 3doh\n│   └── 3dpw\n└── pare\n    └── checkpoints\n\n```\n\nThen, you can evaluate PARE by running:\n\n```shell script\npython scripts/eval.py \\\n  --cfg data/pare/checkpoints/pare_config.yaml \\\n  --opts DATASET.VAL_DS 3doh_3dpw-all\n  \npython scripts/eval.py \\\n  --cfg data/pare/checkpoints/pare_w_3dpw_config.yaml \\\n  --opts DATASET.VAL_DS 3doh_3dpw-all\n```\n\nYou should obtain results in this table on 3DPW test set:\n\n| | MPJPE | PAMPJPE | V2V|\n|--- | --- | --- | ---|\n|PARE | 82 | 50.9 | 97.9|\n|PARE (w. 3DPW) | 74.5 | 46.5 | 88.6|\n\n## Occlusion Sensitivity Analysis\n\nWe prepare a script to run occlusion sensitivity analysis\nproposed in our paper. Occlusion sensitivity analysis slides\nan occluding patch on the image and visualizes how human pose\nand shape estimation result affected.\n\n```shell\npython scripts/occlusion_analysis.py \\\n  --cfg data/pare/checkpoints/pare_config.yaml \\\n  --ckpt data/pare/checkpoints/pare_checkpoint.ckpt\n```\n\nSample occlusion test output:\n\n\u003cp float=\"left\"\u003e\n  \u003cimg src=\"docs/assets/occlusion_test.gif\" width=\"70%\" /\u003e\n\u003c/p\u003e\n\n## Citation\n\n```bibtex\n@inproceedings{Kocabas_PARE_2021,\n  title = {{PARE}: Part Attention Regressor for {3D} Human Body Estimation},\n  author = {Kocabas, Muhammed and Huang, Chun-Hao P. and Hilliges, Otmar and Black, Michael J.},\n  booktitle = {Proc. International Conference on Computer Vision (ICCV)},\n  pages = {11127--11137},\n  month = oct,\n  year = {2021},\n  doi = {},\n  month_numeric = {10}\n}\n```\n## License\n\nThis code is available for **non-commercial scientific research purposes** as defined in the [LICENSE file](LICENSE). By downloading and using this code you agree to the terms in the [LICENSE](LICENSE). Third-party datasets and software are subject to their respective licenses.\n\n## References\n\nWe indicate if a function or script is borrowed externally inside each file. Consider citing these works if you use them in your project.\n\n## Contact\n\nFor questions, please contact pare@tue.mpg.de\n\nFor commercial licensing (and all related questions for business applications), please contact ps-licensing@tue.mpg.de.\n\n\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fmkocabas%2Fpare","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fmkocabas%2Fpare","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fmkocabas%2Fpare/lists"}