{"id":13935706,"url":"https://github.com/JDAI-CV/centerX","last_synced_at":"2025-07-19T20:33:47.350Z","repository":{"id":43462697,"uuid":"317481466","full_name":"JDAI-CV/centerX","owner":"JDAI-CV","description":"This repo is implemented based on detectron2 and centernet","archived":false,"fork":false,"pushed_at":"2020-12-31T08:34:52.000Z","size":196,"stargazers_count":555,"open_issues_count":16,"forks_count":86,"subscribers_count":18,"default_branch":"master","last_synced_at":"2024-11-23T17:04:20.557Z","etag":null,"topics":["caffe","centernet","centerx","deep-learning","detectron2","fast-reid","object-detection","onnx","tensorrt"],"latest_commit_sha":null,"homepage":"","language":"Python","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"apache-2.0","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/JDAI-CV.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}},"created_at":"2020-12-01T08:56:03.000Z","updated_at":"2024-11-12T16:28:40.000Z","dependencies_parsed_at":"2022-08-12T10:41:35.386Z","dependency_job_id":null,"html_url":"https://github.com/JDAI-CV/centerX","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/JDAI-CV%2FcenterX","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/JDAI-CV%2FcenterX/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/JDAI-CV%2FcenterX/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/JDAI-CV%2FcenterX/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/JDAI-CV","download_url":"https://codeload.github.com/JDAI-CV/centerX/tar.gz/refs/heads/master","host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":226677295,"owners_count":17666028,"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":["caffe","centernet","centerx","deep-learning","detectron2","fast-reid","object-detection","onnx","tensorrt"],"created_at":"2024-08-07T23:02:00.881Z","updated_at":"2024-11-27T03:31:12.206Z","avatar_url":"https://github.com/JDAI-CV.png","language":"Python","funding_links":[],"categories":["Python"],"sub_categories":[],"readme":"\u003cimg src=\"cplogo.png\" width=\"300\" \u003e\n\n## \n\nThis repo is implemented based on [detectron2](https://github.com/facebookresearch/detectron2) and  [CenterNet](https://github.com/xingyizhou/CenterNet)\n\n## What\\'s new\n- Support [imgaug](https://github.com/aleju/imgaug.git) data augmentation\n- Support [swa](https://pytorch.org/blog/stochastic-weight-averaging-in-pytorch/)\n- Support **Knowledge Distill**, teacher-student, designed by myself\n- Support other LR_SCHEDULER\n- Support Optimizer [RangerLars](https://github.com/pabloppp/pytorch-tools.git), not convergence in COCO\n- We provide some examples and scripts to convert centerX to Caffe, ONNX and TensorRT format in [projects/speedup](https://github.com/CPFLAME/centerX/tree/master/projects/speedup)\n \n## What\\'s comming \n- [️✔] Support simple inference \n- [✔] Support to caffe, onnx, tensorRT\n- [ ] Support keypoints \n\n## Requirements\n\n- Python \u003e= 3.7\n- PyTorch \u003e= 1.5\n- torchvision that matches the PyTorch installation.\n- OpenCV\n- pycocotools\n\n```shell\npip install cython; pip install 'git+https://github.com/cocodataset/cocoapi.git#subdirectory=PythonAPI'\n```\n\n- GCC \u003e= 4.9\n\n```shell\ngcc --version\n```\n\n- detectron2\n\n```shell\npip install -U 'git+https://github.com/CPFLAME/detectron2.git'\n```\n\n- [pytorch tools](https://github.com/pabloppp/pytorch-tools.git)\n\n```shell\npip install git+https://github.com/pabloppp/pytorch-tools -U\n```\n### Data prepare\nthe same as [detectron2](https://detectron2.readthedocs.io/tutorials/builtin_datasets.html)\n\n### Training\n\nmodify your yamls in run.sh\n```shell\nsh run.sh\n```\n\n### Testing and Evaluation\n\nmodify your yamls in run.sh, \nadd eval-only and MODEL.WEIGHTS in your setting\n```shell\nsh run.sh\n```\n\n## Performance\n\n### coco\n\nThis repo use less training time to get a competitive performance compared to other versions\n\nBackbone ResNet-50\n\n| Code             | mAP  | epoch |\n| ---------------- | ---- | ----- |\n| centerX          | 33.2 |  70   |\n| centerX          | 34.3 |  140  |\n| centernet-better | 34.9 |  140  |\n\nBackbone ResNet-18\n\ncenterX_KD means ResNet-50(33.2) as teacher, ResNet-18(27.9) as student, Knowledge Distill for 70 epoch in coco.\n\n| Code             | mAP  | epoch |\n| ---------------- | ---- | ----- |\n| centerX          | 30.2 |  140  |\n| centerX          | 27.9 |  70   |\n| centerX_KD       | 31.0 |  70   |\n| centernet-better | 29.8 |  140  |\n| centernet        | 28.1 |  140  |\n\n### crowd human\n- optim: SGD\n- lr: 0.02\n- scheduler: WarmupMultiStepLR ,drop 0.1 in (50, 62) for 80 epoch; (90 ,120) for 140 epoch \n- train size: 512 max size\n- test size: 512 max size\n- batch size: 64\n- woGT: KD only use teacher loss\n\n\n| Backbone                 | mAP  | mAP50 | mAP75 | epoch | teacher | student_pretrain |\n| ----------------         | ---- | ----- | ----- | ----- |  -----  | ------- |\n| resdcn18                 | 31.2 | 56.6  | 30.8  |  80   |   -     |  -      |\n| resdcn18_swa             | 31.1 | 56.6  | 30.4  |  80   |   -     |  -      |\n| resdcn18_syncBN          | 31.3 | 56.6  | 30.7  |  80   |   -     |  -      |\n| resdcn18_imgaug          | 29.6 | 54.7  | 28.9  |  80   |   -     |  -      |\n| resdcn18_KD              | 34.5 | 60.2  | 34.3  |  80   | resdcn50| resdcn18|\n| resdcn18_KD_woGT         | 33.0 | 58.3  | 32.7  |  80   | resdcn50| resdcn18|\n| resdcn18_KD_woGT_scratch | 32.8 | 58.1  | 32.6  |  140  | resdcn50| imagenet|\n| resdcn50                 | 35.1 | 61.2  | 35.3  |  80   |   -     |  -      |\n \n## KD exp\n\n### crowd human KD\nGeneralization performance for Knowledge Distill\n\n| Backbone                 | crowd mAP  | coco_person mAP  | epoch | teacher | student_pretrain | train_set |\n| ----------------         | ----       | --------------   | ----- |  -----  | -------          | -----     |\n| resdcn50                 | **35.1**   |    35.7          |  80   |   -     |    -             | crowd     |\n| resdcn18(baseline)       | 31.2       |    31.2          |  80   |   -     |    -             | crowd     |\n| resdcn18_KD              | 34.5       |    34.9          |  80   | resdcn50| resdcn18         | crowd     |\n| resdcn18_KD_woGT_scratch | 32.8       |    34.2          |  140  | resdcn50| imagenet         | crowd     |\n| resdcn18_KD_woGT_scratch | 34.1       |  **36.3**        |  140  | resdcn50| imagenet         | crowd+coco|\n\n### multi teacher KD\n\n| Backbone                 |  mAP crowd     |  mAP coco_car  | epoch | teacher | student_pretrain | train_set |\n| ----------------         | ----           | ----           | ----- |  -----  | -------          | -----     |\n| 1.resdcn50               | 35.1           | -              |  80   |   -     |    -             | crowd     |\n| 2.resdcn18               | 31.7           | -              |  70   |   -     |    -             | crowd     |\n| 3.resdcn50               | -              | 31.6           |  70   |   -     |    -             | coco_car  |\n| 4.resdcn18               | -              | 27.8           |  70   |   -     |    -             | coco_car  |\n| resdcn18_KD_woGT_scratch | 31.6           | 29.4           |  140  |  1,3    | imagenet         | crowd+coco_car|\n\n| Backbone                 |  mAP crowd_human |  mAP widerface | epoch | teacher | student_pretrain | train_set |\n| ----------------         | ----             | -------------- | ----- |  -----  | -------          | -----     |\n| 1.resdcn50               | 35.1             | -              |  80   |   -     |    -             | crowd     |\n| 2.resdcn18               | 31.7             | -              |  70   |   -     |    -             | crowd     |\n| 3.resdcn50               | -                | 32.9           |  70   |   -     |    -             | widerface |\n| 4.resdcn18               | -                | 29.6           |  70   |   -     |    -             | widerface |\n| 5.resdcn18_ignore_nolabel| 29.1             | 24.2           |  140   |   -     |    -             |crowd+wider|\n| 6.resdcn18_pseudo_label  | 28.9             | 27.7           |  140   |   -     |    -             |crowd+wider|\n| 7.resdcn18_KD_woGT_scratch| 31.3            | 32.1           |  140  |  1,3    | imagenet         |crowd+wider|\n\n## License\n\ncenterX is released under the [Apache 2.0 license.](https://github.com/CPFLAME/centerX/blob/master/LICENSE)\n\n## Acknowledgement\n\n- [detectron2](https://github.com/facebookresearch/detectron2)\n- [CenterNet](https://github.com/xingyizhou/CenterNet)\n- [CenterNet-better](https://github.com/FateScript/CenterNet-better)\n- [CenterNet-better-plus](https://github.com/lbin/CenterNet-better-plus.git)\n- [FastReID](https://github.com/JDAI-CV/fast-reid)\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2FJDAI-CV%2FcenterX","html_url":"https://awesome.ecosyste.ms/projects/github.com%2FJDAI-CV%2FcenterX","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2FJDAI-CV%2FcenterX/lists"}