{"id":13676504,"url":"https://github.com/happywu/Sequence-Level-Semantics-Aggregation","last_synced_at":"2025-04-29T07:32:31.399Z","repository":{"id":99800401,"uuid":"215213373","full_name":"happywu/Sequence-Level-Semantics-Aggregation","owner":"happywu","description":"Sequence Level Semantics Aggregation for Video Object Detection","archived":false,"fork":false,"pushed_at":"2024-08-30T23:42:49.000Z","size":3681,"stargazers_count":85,"open_issues_count":6,"forks_count":19,"subscribers_count":4,"default_branch":"master","last_synced_at":"2024-11-11T18:42:55.224Z","etag":null,"topics":["mxnet","object-detection","video-object-detection"],"latest_commit_sha":null,"homepage":null,"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/happywu.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":"2019-10-15T05:23:36.000Z","updated_at":"2024-05-23T09:39:46.000Z","dependencies_parsed_at":"2023-08-01T01:32:08.979Z","dependency_job_id":null,"html_url":"https://github.com/happywu/Sequence-Level-Semantics-Aggregation","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/happywu%2FSequence-Level-Semantics-Aggregation","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/happywu%2FSequence-Level-Semantics-Aggregation/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/happywu%2FSequence-Level-Semantics-Aggregation/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/happywu%2FSequence-Level-Semantics-Aggregation/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/happywu","download_url":"https://codeload.github.com/happywu/Sequence-Level-Semantics-Aggregation/tar.gz/refs/heads/master","host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":251456004,"owners_count":21592273,"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":["mxnet","object-detection","video-object-detection"],"created_at":"2024-08-02T13:00:28.626Z","updated_at":"2025-04-29T07:32:31.392Z","avatar_url":"https://github.com/happywu.png","language":"Python","funding_links":[],"categories":["Frameworks"],"sub_categories":[],"readme":"# Sequence Level Semantics Aggregation for Video Object Detection\r\n\r\n## Introduction\r\nThis is an official MXNet implementation of \r\n[*Sequence Level Semantics Aggregation for Video Object Detection*](https://arxiv.org/abs/1907.06390). (ICCV 2019, oral).\r\nSELSA aggregates full-sequence level information of videos while keeping a simple and clean pipeline. It achieves **82.69**\r\nmAP with ResNet-101 on ImageNet VID validation set.\r\n\r\n## Citation\r\nIf you use the code or models in your research, please cite with:\r\n```\r\n@article{wu2019selsa,\r\n  title={Sequence Level Semantics Aggregation for Video Object Detection},\r\n  author={Wu, Haiping and Chen, Yuntao and Wang, Naiyan and Zhang, Zhaoxiang},\r\n  journal={ICCV 2019},\r\n  year={2019}\r\n}\r\n```\r\n\r\n## Main Results\r\n\r\n\r\n|                                 | \u003csub\u003etraining data\u003c/sub\u003e  | \u003csub\u003etesting data\u003c/sub\u003e | \u003csub\u003emAP(%)\u003c/sub\u003e | \u003csub\u003emAP(%)\u003c/br\u003e(slow)\u003c/sub\u003e  | \u003csub\u003emAP(%)\u003c/br\u003e(medium)\u003c/sub\u003e | \u003csub\u003emAP(%)\u003c/br\u003e(fast)\u003c/sub\u003e |\r\n|---------------------------------|-------------------|--------------|---------|---------|--------|--------|\r\n| \u003csub\u003eSingle-frame baseline\u003c/br\u003e(Faster R-CNN, ResNet-101)\u003c/sub\u003e   | \u003csub\u003eImageNet DET train\u003c/br\u003e + VID train\u003c/sub\u003e | \u003csub\u003eImageNet VID validation\u003c/sub\u003e | 73.6 | 82.1 | 71.0 | 52.5 |\r\n| \u003csub\u003eSELSA\u003c/br\u003e(Faster R-CNN, ResNet-101)\u003c/sub\u003e  | \u003csub\u003eImageNet DET train\u003c/br\u003e + VID train\u003c/sub\u003e | \u003csub\u003eImageNet VID validation\u003c/sub\u003e | 80.3| 86.9 | 78.9 | 61.4 |\r\n| \u003csub\u003eSELSA\u003c/br\u003e(Faster R-CNN, ResNet-101, Data Aug)\u003c/sub\u003e  | \u003csub\u003eImageNet DET train\u003c/br\u003e + VID train\u003c/sub\u003e | \u003csub\u003eImageNet VID validation\u003c/sub\u003e | 82.7 | 88.0 | 81.4 | 67.1 |\r\n\r\n\r\n\r\n## Installation\r\n\r\nPlease note that this repo is based on Python 2.\r\n\r\n1. Clone the repository.\r\n~~~\r\ngit clone https://github.com/happywu/Sequence-Level-Semantics-Aggregation\r\n~~~\r\n\r\n2. Install MXNet following https://mxnet.incubator.apache.org/get_started. We tested our code on MXNet v1.3.0.\r\n\r\n3. Install packages via \r\n~~~\r\npip install -r requirements.txt\r\nsh init.sh\r\n~~~\r\n\r\n## Preparation for Training \u0026 Testing\r\n\r\n1. Please download ILSVRC2015 DET and ILSVRC2015 VID dataset, and make sure it looks like this:\r\n\r\n\t```\r\n\t./data/ILSVRC2015/\r\n\t./data/ILSVRC2015/Annotations/DET\r\n\t./data/ILSVRC2015/Annotations/VID\r\n\t./data/ILSVRC2015/Data/DET\r\n\t./data/ILSVRC2015/Data/VID\r\n\t./data/ILSVRC2015/ImageSets\r\n\t```\r\n\r\n2. Please download ImageNet pre-trained [ResNet-v1-101](https://1dv.aflat.top/resnet_v1_101-0000.params) model and \r\nour pretrained [SELSA ResNet-101](https://1dv.aflat.top/selsa_rcnn_vid-0000.params) model manually, and put it under folder `./model`. Make sure it looks like this:\r\n\t```\r\n\t./model/pretrained_model/resnet_v1_101-0000.params\r\n\t./model/pretrained_model/selsa_rcnn_vid-0000.params\r\n\t```\r\n## Testing\r\n1. To test the provided pretrained model, run the following command.\r\n    ```\r\n    python experiments/selsa/test.py --cfg experiments/selsa/cfgs/resnet_v1_101_rcnn_selsa_aug.yaml --test-pretrained ./model/pretrained_model/selsa_rcnn_vid\r\n    ```\r\n   \r\nYou should get the results as reported before.\r\n## Training\r\n\r\n3. To train, use the following command\r\n    ```\r\n    python experiments/selsa/train_end2end.py --cfg experiments/selsa/cfgs/resnet_v1_101_rcnn_selsa_aug.yaml\r\n    ```\r\n\tA cache folder would be created automatically to save the model and the log under `output/selsa_rcnn/imagenet_vid/`.\r\n\t\r\n2. To test your trained model\r\n    ```\r\n    python experiments/selsa/test.py --cfg experiments/selsa/cfgs/resnet_v1_101_rcnn_selsa_aug.yaml\r\n    ```\r\n\r\n## Other implementations\r\nPytorch: [MMTracking](https://github.com/open-mmlab/mmtracking/tree/master/configs/vid/selsa)\r\n\r\n## Acknowledge\r\nThis repo is modified from [*Flow-Guided-Feature-Aggregation*](https://github.com/msracver/Flow-Guided-Feature-Aggregation).\r\n\r\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fhappywu%2FSequence-Level-Semantics-Aggregation","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fhappywu%2FSequence-Level-Semantics-Aggregation","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fhappywu%2FSequence-Level-Semantics-Aggregation/lists"}