{"id":19065638,"url":"https://github.com/xingyizhou/unidet","last_synced_at":"2025-04-13T04:59:32.333Z","repository":{"id":45019085,"uuid":"342336290","full_name":"xingyizhou/UniDet","owner":"xingyizhou","description":"Object detection on multiple datasets with an automatically learned unified label space.","archived":false,"fork":false,"pushed_at":"2024-03-08T07:05:47.000Z","size":9218,"stargazers_count":505,"open_issues_count":24,"forks_count":56,"subscribers_count":13,"default_branch":"master","last_synced_at":"2025-04-13T04:59:22.008Z","etag":null,"topics":["coco","object-detection","objects365","openimages","robust"],"latest_commit_sha":null,"homepage":"","language":"Python","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/xingyizhou.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":"2021-02-25T18:11:13.000Z","updated_at":"2025-04-09T07:20:42.000Z","dependencies_parsed_at":"2024-11-09T01:04:31.094Z","dependency_job_id":null,"html_url":"https://github.com/xingyizhou/UniDet","commit_stats":null,"previous_names":[],"tags_count":7,"template":false,"template_full_name":null,"repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/xingyizhou%2FUniDet","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/xingyizhou%2FUniDet/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/xingyizhou%2FUniDet/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/xingyizhou%2FUniDet/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/xingyizhou","download_url":"https://codeload.github.com/xingyizhou/UniDet/tar.gz/refs/heads/master","host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":248665757,"owners_count":21142123,"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":["coco","object-detection","objects365","openimages","robust"],"created_at":"2024-11-09T00:51:44.482Z","updated_at":"2025-04-13T04:59:32.292Z","avatar_url":"https://github.com/xingyizhou.png","language":"Python","funding_links":[],"categories":[],"sub_categories":[],"readme":"# Simple multi-dataset detection\nAn object detector trained on multiple large-scale datasets with a unified label space; Winning solution of ECCV 2020 Robust Vision Challenges.\n\n\u003cp align=\"center\"\u003e \u003cimg src='docs/unidet_teaser.jpg' align=\"center\" height=\"170px\"\u003e \u003c/p\u003e\n\n\u003e [**Simple multi-dataset detection**](http://arxiv.org/abs/2102.13086),            \n\u003e Xingyi Zhou, Vladlen Koltun, Philipp Kr\u0026auml;henb\u0026uuml;hl,        \n\u003e *CVPR 2022 ([arXiv 2102.13086](http://arxiv.org/abs/2102.13086))*         \n\nContact: [zhouxy@cs.utexas.edu](mailto:zhouxy@cs.utexas.edu). Any questions or discussions are welcomed! \n\n## Features at a glance\n\n- We trained a unified object detector on 4 large-scale detection datasets: COCO, Objects365, OpenImages, and Mapillary, with state-of-the-art performance on all of them.\n\n- The model predicts class labels in a **learned** unified label space.\n\n- The model can be directly used to test on novel datasets outside the training datasets.\n\n- In this repo, we also provide state-of-the-art baselines for Objects365 and OpenImages.\n\n## Main results\n\n- [RVC challenge](http://www.robustvision.net/leaderboard.php?benchmark=object)\n\n| COCO test-challenge | OpenImages public test | Mapillary test | Objects365 val |\n|---------------------|------------------------|----------------|----------------|\n| 52.9                | 60.6                   | 25.3           | 33.7           |\n\nResults are obtained using a Cascade-RCNN with ResNeSt200 trained in an 8x schedule.\n\n\n\n- Unified model vs. ensemble of dataset-specific models with known test domains.\n\n|                       |  COCO     | Objects365   |  OpenImages  |  mean. |\n|-----------------------|-----------|--------------|--------------|--------|\n|Unified                | 45.4      | 24.4         | 66.0         | 45.3   |\n|Dataset-specific models| 42.5      | 24.9         | 65.7         | 44.4   |\n\nResults are obtained using a Cascade-RCNN with Res50 trained in an 8x schedule.\n\n- Zero-shot cross dataset evaluation\n\n|                |  VOC  | VIPER |  CityScapes  | ScanNet | WildDash | CrowdHuman | KITTI | mean |\n|----------------|-------|-------|--------------|---------|----------|------------|-------|------|\n|Unified         | 82.9  | 21.3  | 52.6         | 29.8    | 34.7     | 70.7       | 39.9  | 47.3 |\n|Oracle models   | 80.3  | 31.8  | 54.6         | 44.7    | -        | 80.0       | -     | -    |\n\nResults are obtained using a Cascade-RCNN with Res50 trained in an 8x schedule.\n\nMore models can be found in our [MODEL ZOO](docs/REPRODUCE.md).\n\n## Installation\n\nOur project is developed on [detectron2](https://github.com/facebookresearch/detectron2). Please follow the official [detectron2 installation](https://github.com/facebookresearch/detectron2/blob/master/INSTALL.md).\n\n## Demo\n\nWe use the same inference API as detectorn2. To run inference on an image folder using our pretrained model, run\n\n~~~\npython demo.py --config-file configs/Unified_learned_OCIM_R50_6x+2x.yaml --input images/*.jpg --opts MODEL.WEIGHTS models/Unified_learned_OCIM_R50_6x+2x.pth\n~~~\n\nIf setup correctly, the output should look like:\n\u003cp align=\"center\"\u003e \u003cimg src='docs/example_output2.jpg' align=\"center\" height=\"460px\"\u003e \u003c/p\u003e\n\n*The sample image is from [WildDash](https://wilddash.cc/) dataset.\n\nNote that the model predicts all labels in its label hierarchy tree (for example, both `vehicle` and `car` for a car), following the protocol in OpenImages.\n\n## Benchmark evaluation and training\n\nAfter installation, follow the instructions in [DATASETS.md](docs/DATASETS.md) to setup the (many) datasets. Then check [REPRODUCE.md](docs/REPRODUCE.md) to reproduce the results in the paper.\n\n## License\n\nOur code is under [Apache 2.0 license](LICENSE).\n\n\n## Citation\n\nIf you find this project useful for your research, please use the following BibTeX entry.\n\n    @inproceedings{zhou2021simple,\n      title={Simple multi-dataset detection},\n      author={Zhou, Xingyi and Koltun, Vladlen and Kr{\\\"a}henb{\\\"u}hl, Philipp},\n      booktitle={CVPR},\n      year={2022}\n    }","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fxingyizhou%2Funidet","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fxingyizhou%2Funidet","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fxingyizhou%2Funidet/lists"}