{"id":16393954,"url":"https://github.com/jingkang50/openpsg","last_synced_at":"2025-04-04T09:09:32.750Z","repository":{"id":46168958,"uuid":"515119088","full_name":"Jingkang50/OpenPSG","owner":"Jingkang50","description":"Benchmarking Panoptic Scene Graph Generation (PSG), ECCV'22","archived":false,"fork":false,"pushed_at":"2023-04-10T03:50:49.000Z","size":6527,"stargazers_count":414,"open_issues_count":26,"forks_count":68,"subscribers_count":6,"default_branch":"main","last_synced_at":"2024-10-12T04:54:37.972Z","etag":null,"topics":["scene-graph","scene-graph-generation","scene-understanding"],"latest_commit_sha":null,"homepage":"https://psgdataset.org","language":"Python","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/Jingkang50.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":"CONTRIBUTING.md","funding":null,"license":"LICENSE","code_of_conduct":"CODE_OF_CONDUCT.md","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":"2022-07-18T09:34:35.000Z","updated_at":"2024-10-07T15:02:27.000Z","dependencies_parsed_at":"2024-10-25T17:56:53.832Z","dependency_job_id":"95b7e379-64dd-48e1-9c2a-c5a713dd22da","html_url":"https://github.com/Jingkang50/OpenPSG","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/Jingkang50%2FOpenPSG","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/Jingkang50%2FOpenPSG/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/Jingkang50%2FOpenPSG/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/Jingkang50%2FOpenPSG/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/Jingkang50","download_url":"https://codeload.github.com/Jingkang50/OpenPSG/tar.gz/refs/heads/main","host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":247149502,"owners_count":20891954,"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":["scene-graph","scene-graph-generation","scene-understanding"],"created_at":"2024-10-11T04:54:35.144Z","updated_at":"2025-04-04T09:09:32.730Z","avatar_url":"https://github.com/Jingkang50.png","language":"Python","funding_links":[],"categories":[],"sub_categories":[],"readme":"# Panoptic Scene Graph Generation\n\u003c!-- \u003cbr /\u003e --\u003e\n\u003c!-- \u003cp align=\"center\"\u003e\n  \u003cimg src=\"https://live.staticflickr.com/65535/52193879677_751a4e0b79_k.jpg\" align=\"center\" width=\"60%\"\u003e --\u003e\n\u003cp align=\"center\"\u003e\n  \u003cimg src=\"./assets/psgtr_long.gif\" align=\"center\" width=\"80%\"\u003e\n\n  \u003cp align=\"center\"\u003e\n  \u003ca href=\"https://arxiv.org/abs/2207.11247\" target='_blank'\u003e\n    \u003cimg src=\"https://img.shields.io/badge/Paper-ECCV%202022-b31b1b?style=flat-square\"\u003e\n  \u003c/a\u003e\n  \u0026nbsp;\u0026nbsp;\u0026nbsp;\n  \u003ca href=\"https://psgdataset.org/\" target='_blank'\u003e\n    \u003cimg src=\"https://img.shields.io/badge/Page-psgdataset.org-228c22?style=flat-square\"\u003e\n  \u003c/a\u003e\n  \u0026nbsp;\u0026nbsp;\u0026nbsp;\n  \u003ca href=\"https://entuedu-my.sharepoint.com/:f:/g/personal/jingkang001_e_ntu_edu_sg/EgQzvsYo3t9BpxgMZ6VHaEMBDAb7v0UgI8iIAExQUJq62Q?e=fIY3zh\" target='_blank'\u003e\n    \u003cimg src=\"https://img.shields.io/badge/Data-PSGDataset-334b7f?style=flat-square\"\u003e\n  \u003c/a\u003e\n  \u0026nbsp;\u0026nbsp;\u0026nbsp;\n  \u003ca href=\"https://www.cvmart.net/race/10349/base\" target='_blank'\u003e\n    \u003cimg src=\"https://img.shields.io/badge/Competition-PSG Challenge-f2d297?style=flat-square\"\u003e\n  \u003c/a\u003e\n  \u003cbr\u003e\n  \u003ca href=\"https://huggingface.co/spaces/mmlab-ntu/OpenPSG\" target='_blank'\u003e\n    \u003cimg src=\"https://img.shields.io/badge/Demo-HuggingFace-ffca37?style=flat-square\"\u003e\n  \u003c/a\u003e\n  \u0026nbsp;\u0026nbsp;\u0026nbsp;\n  \u003ca href=\"https://paperswithcode.com/task/panoptic-scene-graph-generation/\" target='_blank'\u003e\n    \u003cimg src=\"https://img.shields.io/badge/Benchmark-PapersWithCode-00c4c6?style=flat-square\"\u003e\n  \u003c/a\u003e\n  \u0026nbsp;\u0026nbsp;\u0026nbsp;\n  \u003ca href=\"https://join.slack.com/t/psgdataset/shared_invite/zt-1f8wkjfky-~uikum1YA1giLGZphFZdAQ\" target='_blank'\u003e\n    \u003cimg src=\"https://img.shields.io/badge/Forum-Slack-4c1448?style=flat-square\"\u003e\n    \u0026nbsp;\u0026nbsp;\u0026nbsp;\n  \u003ca href=\"https://replicate.com/cjwbw/openpsg\" target='_blank'\u003e\n    \u003cimg src=\"https://img.shields.io/badge/Replicate-Demo \u0026 Cloud API-1b82c2?style=flat-square\"\u003e\n  \u003c/a\u003e\n\n\n\u003c/p\u003e\n\n\n  \u003cp align=\"center\"\u003e\n  \u003cfont size=5\u003e\u003cstrong\u003ePanoptic Scene Graph Generation\u003c/strong\u003e\u003c/font\u003e\n    \u003cbr\u003e\n      \u003ca href=\"http://jingkang50.github.io/\" target='_blank'\u003eJingkang Yang\u003c/a\u003e,\u0026nbsp;\n      \u003ca href=\"https://yizhe-ang.github.io/\" target='_blank'\u003eYi Zhe Ang\u003c/a\u003e,\u0026nbsp;\n      \u003ca href=\"https://www.linkedin.com/in/zujin-guo-652b0417a/\" target='_blank'\u003eZujin Guo\u003c/a\u003e,\u0026nbsp;\n      \u003ca href=\"https://kaiyangzhou.github.io/\" target='_blank'\u003eKaiyang Zhou\u003c/a\u003e,\u0026nbsp;\n      \u003ca href=\"http://www.statfe.com/\" target='_blank'\u003eWayne Zhang\u003c/a\u003e,\u0026nbsp;\n      \u003ca href=\"https://liuziwei7.github.io/\" target='_blank'\u003eZiwei Liu\u003c/a\u003e\n    \u003cbr\u003e\n  S-Lab, Nanyang Technological University \u0026 SenseTime Research\n  \u003c/p\u003e\n\u003c/p\u003e\n\n\n---\n\n## Updates\n- **Oct 31, 2022**: We release the full dataset [here](https://entuedu-my.sharepoint.com/:f:/g/personal/jingkang001_e_ntu_edu_sg/EgQzvsYo3t9BpxgMZ6VHaEMBDAb7v0UgI8iIAExQUJq62Q?e=fIY3zh) that we used in the paper. The competition version is [here](https://entuedu-my.sharepoint.com/:f:/g/personal/jingkang001_e_ntu_edu_sg/Es2o9zfHvD5Fh0MoI9EB6UUBPAsPFEdZtqjA6RWxncUDgA?e=5swsAx).\n- **Oct 9, 2022**: The preliminary round of PSG challenge ends. We will release the entire dataset after the final round starts. Before that, if you want to access the PSG dataset (competition version), please [email](mailto:jingkang001@e.ntu.edu.sg) me.\n- **Sep 4, 2022**: We introduce the PSG Classification Task for NTU CE7454 Coursework, as described [here](https://github.com/Jingkang50/OpenPSG/blob/main/ce7454).\n- **Aug 21, 2022**: We provide guidance on PSG challenge registration [here](https://github.com/Jingkang50/OpenPSG/blob/main/psg_challenge.md).\n- **Aug 12, 2022**: Replicate demo and Cloud API is added, try it [here](https://replicate.com/cjwbw/openpsg)!\n- **Aug 10, 2022**: We launched [Hugging Face demo 🤗](https://huggingface.co/spaces/mmlab-ntu/OpenPSG). Try it with your scene!\n- **Aug 5, 2022**: The PSG Challenge will be available on [International Algorithm Case Competition ](https://iacc.pazhoulab-huangpu.com/)! All the data will be available there then! Stay tuned!\n- **July 25, 2022**: :boom: We are preparing a PSG competition with [ECCV'22 SenseHuman Workshop](https://sense-human.github.io) and [International Algorithm Case Competition](https://iacc.pazhoulab-huangpu.com/), starting from Aug 6, with a prize pool of :money_mouth_face: **US$150K** :money_mouth_face:. Join us on our [Slack](https://join.slack.com/t/psgdataset/shared_invite/zt-1f8wkjfky-~uikum1YA1giLGZphFZdAQ) to stay updated!\n- **July 25, 2022**: PSG paper is available on [arXiv](https://arxiv.org/abs/2207.11247).\n- **July 3, 2022**: PSG is accepted by ECCV'22.\n## What is PSG Task?\n\u003cstrong\u003eThe Panoptic Scene Graph Generation (PSG) Task\u003c/strong\u003e aims to interpret a complex scene image with a scene graph representation, with each node in the scene graph grounded by its pixel-accurate segmentation mask in the image.\n\nTo promote comprehensive scene understanding, we take into account all the content in the image, including \"things\" and \"stuff\", to generate the scene graph.\n\n| ![psg.jpg](https://live.staticflickr.com/65535/52231748332_4945d88929_b.jpg) |\n|:--:|\n| \u003cb\u003ePSG Task: To generate a scene graph that is grounded by its panoptic segmentation\u003c/b\u003e|\n\n\u003c!-- ## Demo of the Current SOTA PSGTR --\u003e\n\n\n\u003c!-- ## Demo of the Current SOTA PSGTR --\u003e\n\n\n## PSG addresses many SGG problems\nWe believe that the biggest problem of classic scene graph generation (SGG) comes from noisy datasets.\nClassic scene graph generation datasets adopt a bounding box-based object grounding, which inevitably causes a number of issues:\n- **Coarse localization**: bounding boxes cannot reach pixel-level accuracy,\n- **Inability to ground comprehensively**: bounding boxes cannot ground backgrounds,\n- **Tendency to provide trivial information**: current datasets usually capture frivolous objects like `head` to form trivial relations like `person-has-head`, due to too much freedom given during bounding box annotation.\n- **Duplicate groundings**: the same object could be grounded by multiple separate bounding boxes.\n\nAll of the problems above can be easily addressed by the PSG dataset, which grounds the objects using panoptic segmentation with an appropriate granularity of object categories (adopted from COCO).\n\nIn fact, the PSG dataset contains 49k overlapping images from COCO and Visual Genome. In a nutshell, we asked annotators to annotate relations based on COCO panoptic segmentations, i.e., relations are mask-to-mask.\n\n| ![psg.jpg](https://live.staticflickr.com/65535/52231743087_2bda038ee2_b.jpg) |\n|:--:|\n| \u003cb\u003eComparison between the classic VG-150 and PSG.\u003c/b\u003e|\n\n## Clear Predicate Definition\nWe also find that a good definition of predicates is unfortunately ignored in the previous SGG datasets.\nTo better formulate PSG task, we carefully define 56 predicates for PSG dataset.\nWe try hard to avoid trivial or duplicated relations, and find that the designed 56 predicates are enough to cover the entire PSG dataset (or common everyday scenarios).\n\nType    | Predicates  |\n---    | ---       |\nPositional Relations (6)     | over, in front of, beside, on, in, attached to. |\nCommon Object-Object Relations (5) | hanging from, on the back of, falling off, going down, painted on.|\nCommon Actions (31) | walking on, running on, crossing, standing on, lying on, sitting on, leaning on, flying over, jumping over, jumping from, wearing, holding, carrying, looking at, guiding, kissing, eating, drinking, feeding, biting, catching, picking (grabbing), playing with, chasing, climbing, cleaning (washing, brushing), playing, touching, pushing, pulling, opening.|\nHuman Actions (4)\t | cooking, talking to, throwing (tossing), slicing.\nActions in Traffic Scene (4) |\tdriving, riding, parked on, driving on.\nActions in Sports Scene (3)\t| about to hit, kicking, swinging.\nInteraction between Background (3) |\tentering, exiting, enclosing (surrounding, warping in)\n\n\n## Get Started\nTo setup the environment, we use `conda` to manage our dependencies.\n\nOur developers use `CUDA 10.1` to do experiments.\n\nYou can specify the appropriate `cudatoolkit` version to install on your machine in the `environment.yml` file, and then run the following to create the `conda` environment:\n```bash\nconda env create -f environment.yml\n```\nYou shall manually install the following dependencies.\n```bash\n# Install mmcv\n## CAUTION: The latest versions of mmcv 1.5.3, mmdet 2.25.0 are not well supported, due to bugs in mmdet.\npip install mmcv-full==1.4.3 -f https://download.openmmlab.com/mmcv/dist/cu101/torch1.7.0/index.html\n\n# Install mmdet\npip install openmim\nmim install mmdet==2.20.0\n\n# Install coco panopticapi\npip install git+https://github.com/cocodataset/panopticapi.git\n\n# For visualization\nconda install -c conda-forge pycocotools\npip install detectron2==0.5 -f \\\n  https://dl.fbaipublicfiles.com/detectron2/wheels/cu101/torch1.7/index.html\n\n# If you're using wandb for logging\npip install wandb\nwandb login\n\n# If you develop and run openpsg directly, install it from source:\npip install -v -e .\n# \"-v\" means verbose, or more output\n# \"-e\" means installing a project in editable mode,\n# thus any local modifications made to the code will take effect without reinstallation.\n```\n\n[Datasets](https://entuedu-my.sharepoint.com/:f:/g/personal/jingkang001_e_ntu_edu_sg/EgQzvsYo3t9BpxgMZ6VHaEMBDAb7v0UgI8iIAExQUJq62Q?e=fIY3zh) and [pretrained models](https://entuedu-my.sharepoint.com/:f:/g/personal/jingkang001_e_ntu_edu_sg/ErQ4stbMxp1NqP8MF8YPFG8BG-mt5geOrrJfAkeitjzASw?e=9taAaU) are provided. Please unzip the files if necessary.\n\n**Before October 2022, we only release part of the PSG data for competition, where part of the test set annotations are wiped out. Users should change the `json` filename in [`psg.py` (Line 4-5)](https://github.com/Jingkang50/OpenPSG/blob/d66dfa70429001ad80c2a8984be9d86a9da703bc/configs/_base_/datasets/psg.py#L4) to a correct filename for training or submission.**\n\n**For the PSG competition, we provide `psg_train_val.json` (45697 training data + 1000 validation data with GT). Participant should use `psg_val_test.json` (1000 validation data with GT + 1177 test data without GT) to submit. Example submit script is [here](https://github.com/Jingkang50/OpenPSG/blob/main/scripts/imp/submit_panoptic_fpn_r50_sgdet.sh). You can use [`grade.sh`](https://github.com/Jingkang50/OpenPSG/blob/main/scripts/grade.sh) to simulate the competition's grading mechanism locally.**\n\nOur codebase accesses the datasets from `./data/` and pretrained models from `./work_dirs/checkpoints/` by default.\n\nIf you want to play with VG, please download the VG dataset [here](https://entuedu-my.sharepoint.com/:f:/g/personal/jingkang001_e_ntu_edu_sg/EiBEV1Z3ueBJqJVO4j7z0YwBt_Jvj2AqYTRsiIs-8pZowg?e=C2O5yg), and put it into `./data` dir.\nWe have pipeline [here](https://github.com/Jingkang50/OpenPSG/blob/main/openpsg/datasets/sg.py) to process the dataset.\n\n```\n├── ...\n├── configs\n├── data\n│   ├── coco\n│   │   ├── panoptic_train2017\n│   │   ├── panoptic_val2017\n│   │   ├── train2017\n│   │   └── val2017\n│   └── psg\n│       ├── psg_train_val.json\n│       ├── psg_val_test.json\n│       └── ...\n├── openpsg\n├── scripts\n├── tools\n├── work_dirs\n│   ├── checkpoints\n│   ├── psgtr_r50\n│   └── ...\n├── ...\n```\nWe suggest our users to play with `./tools/Visualize_Dataset.ipynb` to quickly get familiar with PSG dataset.\n\nTo train or test PSG models, please see https://github.com/Jingkang50/OpenPSG/tree/main/scripts for scripts of each method. Some example scripts are below.\n\n**Training**\n```bash\n# Single GPU for two-stage methods, debug mode\nPYTHONPATH='.':$PYTHONPATH \\\npython -m pdb -c continue tools/train.py \\\n  configs/psg/motif_panoptic_fpn_r50_fpn_1x_sgdet_psg.py\n\n# Multiple GPUs for one-stage methods, running mode\nPYTHONPATH='.':$PYTHONPATH \\\npython -m torch.distributed.launch \\\n--nproc_per_node=8 --master_port=29500 \\\n  tools/train.py \\\n  configs/psgformer/psgformer_r50_psg.py \\\n  --gpus 8 \\\n  --launcher pytorch\n```\n\n**Testing**\n```bash\n# sh scripts/imp/test_panoptic_fpn_r50_sgdet.sh\nPYTHONPATH='.':$PYTHONPATH \\\npython tools/test.py \\\n  configs/imp/panoptic_fpn_r50_fpn_1x_sgdet_psg.py \\\n  path/to/checkpoint.pth \\\n  --eval sgdet\n```\n\n**Submitting for PSG Competition**\n```bash\n# sh scripts/imp/submit_panoptic_fpn_r50_sgdet.sh\nPYTHONPATH='.':$PYTHONPATH \\\npython tools/test.py \\\n  configs/imp/panoptic_fpn_r50_fpn_1x_sgdet_psg.py \\\n  path/to/checkpoint.pth \\\n  --submit\n```\n\n## OpenPSG: Benchmarking PSG Task\n### Supported methods (Welcome to Contribute!)\n\n\u003cdetails open\u003e\n\u003csummary\u003e\u003cb\u003eTwo-Stage Methods (4)\u003c/b\u003e\u003c/summary\u003e\n\n\u003e - [x] IMP (CVPR'17)\n\u003e - [x] MOTIFS (CVPR'18)\n\u003e - [x] VCTree (CVPR'19)\n\u003e - [x] GPSNet (CVPR'20)\n\u003c/details\u003e\n\n\u003cdetails open\u003e\n\u003csummary\u003e\u003cb\u003eOne-Stage Methods (2)\u003c/b\u003e\u003c/summary\u003e\n\n\u003e - [x] PSGTR (ECCV'22)\n\u003e - [x] PSGFormer (ECCV'22)\n\u003c/details\u003e\n\n\n### Supported datasets (Welcome to Contribute!)\n\n- [x] VG-150 (IJCV'17)\n- [x] PSG (ECCV'22)\n\n\n## Model Zoo\nMethod    | Backbone | #Epoch | R/mR@20 | R/mR@50 | R/mR@100 | ckpt | SHA256\n---       | ---  | --- | --- | --- |--- |--- |--- |\nIMP       | ResNet-50 | 12 | 16.5 / 6.52 | 18.2 / 7.05 | 18.6 / 7.23 |  [link](https://entuedu-my.sharepoint.com/:f:/g/personal/jingkang001_e_ntu_edu_sg/EiTgJ9q2h3hDpyXSdu6BtlQBHAZNwNaYmcO7SElxhkIFXw?e=8fytHc) |7be2842b6664e2b9ef6c7c05d27fde521e2401ffe67dbb936438c69e98f9783c |\nMOTIFS    | ResNet-50 | 12 | 20.0 / 9.10 | 21.7 / 9.57 | 22.0 / 9.69 |  [link](https://entuedu-my.sharepoint.com/:f:/g/personal/jingkang001_e_ntu_edu_sg/Eh4hvXIspUFKpNa_75qwDoEBJTCIozTLzm49Ste6HaoPow?e=ZdAs6z) | 956471959ca89acae45c9533fb9f9a6544e650b8ea18fe62cdead495b38751b8 |\nVCTree    | ResNet-50 | 12 | 20.6 / 9.70 | 22.1 / 10.2 | 22.5 / 10.2 |  [link](https://entuedu-my.sharepoint.com/:f:/g/personal/jingkang001_e_ntu_edu_sg/EhKfi9kqAd9CnSoHztQIChABeBjBD3hF7DflrNCjlHfh9A?e=lWa1bd) |e5fdac7e6cc8d9af7ae7027f6d0948bf414a4a605ed5db4d82c5d72de55c9b58 |\nGPSNet    | ResNet-50 | 12 | 17.8 / 7.03 | 19.6 / 7.49 | 20.1 / 7.67 |  [link](https://entuedu-my.sharepoint.com/:f:/g/personal/jingkang001_e_ntu_edu_sg/EipIhZgVgx1LuK2RUmjRg2sB8JqxMIS5GnPDHeaYy5GF6A?e=5j53VF) | 98cd7450925eb88fa311a20fce74c96f712e45b7f29857c5cdf9b9dd57f59c51 |\nPSGTR     | ResNet-50 | 60 | 28.4 / 16.6 | 34.4 / 20.8 | 36.3 / 22.1 |  [link](https://entuedu-my.sharepoint.com/:f:/g/personal/jingkang001_e_ntu_edu_sg/Eonc-KwOxg9EmdtGDX6ss-gB35QpKDnN_1KSWOj6U8sZwQ?e=zdqwqP) | 1c4ddcbda74686568b7e6b8145f7f33030407e27e390c37c23206f95c51829ed |\nPSGFormer | ResNet-50 | 60 | 18.0 / 14.8 | 19.6 / 17.0 | 20.1 / 17.6 |  [link](https://entuedu-my.sharepoint.com/:f:/g/personal/jingkang001_e_ntu_edu_sg/EnaJchJzJPtGrkl4k09evPIB5JUkkDZ2tSS9F-Hd-1KYzA?e=9QA8Nc) | 2f0015ce67040fa00b65986f6ce457c4f8cc34720f7e47a656b462b696a013b7 |\n\n---\n## Contributing\nWe appreciate all contributions to improve OpenPSG.\nWe sincerely welcome community users to participate in these projects. Please refer to [CONTRIBUTING.md](https://github.com/Jingkang50/OpenOOD/blob/v0.5/CONTRIBUTING.md) for the contributing guideline.\n\n## Acknowledgements\nOpenPSG is developed based on [MMDetection](https://github.com/open-mmlab/mmdetection). Most of the two-stage SGG implementations refer to [MMSceneGraph](https://github.com/Kenneth-Wong/MMSceneGraph) and [Scene-Graph-Benchmark.pytorch](https://github.com/KaihuaTang/Scene-Graph-Benchmark.pytorch).\nWe sincerely appreciate the efforts of the developers from the previous codebases.\n\n## Citation\nIf you find our repository useful for your research, please consider citing our paper:\n```bibtex\n@inproceedings{yang2022psg,\n    author = {Yang, Jingkang and Ang, Yi Zhe and Guo, Zujin and Zhou, Kaiyang and Zhang, Wayne and Liu, Ziwei},\n    title = {Panoptic Scene Graph Generation},\n    booktitle = {ECCV}\n    year = {2022}\n}\n\n@inproceedings{yang2023pvsg,\n    author = {Yang, Jingkang and Peng, Wenxuan and Li, Xiangtai and Guo, Zujin and Chen, Liangyu and Li, Bo and Ma, Zheng and Zhou, Kaiyang and Zhang, Wayne and Loy, Chen Change and Liu, Ziwei},\n    title = {Panoptic Video Scene Graph Generation},\n    booktitle = {CVPR},\n    year = {2023},\n}\n```\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fjingkang50%2Fopenpsg","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fjingkang50%2Fopenpsg","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fjingkang50%2Fopenpsg/lists"}