{"id":13439897,"url":"https://github.com/megvii-research/Far3D","last_synced_at":"2025-03-20T09:30:58.526Z","repository":{"id":205582535,"uuid":"713317831","full_name":"megvii-research/Far3D","owner":"megvii-research","description":"[AAAI2024] Far3D: Expanding the Horizon for Surround-view 3D Object Detection","archived":false,"fork":false,"pushed_at":"2023-12-13T08:48:19.000Z","size":3410,"stargazers_count":144,"open_issues_count":19,"forks_count":11,"subscribers_count":4,"default_branch":"main","last_synced_at":"2024-10-28T02:17:58.122Z","etag":null,"topics":["3d-object-detection","long-range","multi-view"],"latest_commit_sha":null,"homepage":"","language":"Jupyter Notebook","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/megvii-research.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}},"created_at":"2023-11-02T09:26:13.000Z","updated_at":"2024-10-24T13:12:26.000Z","dependencies_parsed_at":null,"dependency_job_id":"fd9b6dcc-cff4-4973-bb86-9d6af5d9d97a","html_url":"https://github.com/megvii-research/Far3D","commit_stats":null,"previous_names":["megvii-research/far3d"],"tags_count":1,"template":false,"template_full_name":null,"repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/megvii-research%2FFar3D","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/megvii-research%2FFar3D/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/megvii-research%2FFar3D/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/megvii-research%2FFar3D/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/megvii-research","download_url":"https://codeload.github.com/megvii-research/Far3D/tar.gz/refs/heads/main","host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":244585679,"owners_count":20476776,"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-object-detection","long-range","multi-view"],"created_at":"2024-07-31T03:01:17.997Z","updated_at":"2025-03-20T09:30:57.431Z","avatar_url":"https://github.com/megvii-research.png","language":"Jupyter Notebook","funding_links":[],"categories":["Jupyter Notebook"],"sub_categories":[],"readme":"\u003cdiv align=\"center\"\u003e\n\u003ch1\u003eFar3D\u003c/h1\u003e\n\u003ch3\u003e [AAAI2024] Expanding the Horizon for Surround-view 3D Object Detection \u003c/h3\u003e\n\u003c/div\u003e\n\n[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/far3d-expanding-the-horizon-for-surround-view/3d-object-detection-on-nuscenes-camera-only)](https://paperswithcode.com/sota/3d-object-detection-on-nuscenes-camera-only?p=far3d-expanding-the-horizon-for-surround-view)\n[![arXiv](https://img.shields.io/badge/arXiv-Paper-\u003cCOLOR\u003e.svg)](https://arxiv.org/abs/2308.09616)\n\n## Introduction\n\nThis repository is an official implementation of Far3D.\n![](figs/framework.png)\n\nExpanding existing methods directly to cover long distances poses challenges such as heavy computation costs and unstable convergence. \nTo address these limitations, we proposes a novel sparse query-based framework, dubbed Far3D. By utilizing high-quality 2D object priors, we generate 3D adaptive queries that complement the 3D global queries.\nTo efficiently capture discriminative features across different views and scales for long-range objects, we introduce a perspective-aware aggregation module. Additionally, we propose a range-modulated 3D denoising approach to address query error propagation and mitigate convergence issues in long-range tasks.\n\n## News\n- [2023/08/01] We release the paper on [Arxiv]((https://arxiv.org/abs/2308.09616)).  \n- [2023/08/01] Far3D achieves comparable performance (31.6 mAP, 23.9 CDS) on long-range Argoverse2 dataset, as well as achieving SoTA performance (63.5 mAP, 68.7 NDS) on nuScenes Camera Only.\n\n## Getting Started\nOur pipeline follows [StreamPETR](https://github.com/exiawsh/StreamPETR), and you can follow [Get Started](./docs/get_started.md) step by step.\n* If you have used StreamPETR before, it is easy to run Far3D without additional extensive installation.\n\nQuick Train \u0026 Evaluation\n\nTrain the model\n```angular2html\ntools/dist_train.sh projects/configs/far3d.py 8 --work-dir work_dirs/far3d/\n```\nEvaluation\n```angular2html\ntools/dist_test.sh projects/configs/far3d.py work_dirs/far3d/iter_82548.pth 8 --eval bbox\n```\n\n## Results on Argoverse 2 Val Set.\n| Model | Backbone | Input size | mAP | CDS | Config | Download |\n| :---: | :---: | :---: | :---: | :---:| :---: | :---:|\n| BEVStereo | VoV-99 | (960, 640) | 0.146 | 0.104 | -- | -- |\n| SOLOFusion  | VoV-99 | (960, 640) | 0.149 | 0.106 | -- | -- |\n| PETR | VoV-99 | (960, 640) |  0.176 | 0.122 | -- | -- |\n| Sparse4Dv2  | VoV-99 | (960, 640) | 0.189 | 0.134 | -- | -- |\n| StreamPETR | VoV-99 | (960, 640) | 0.203 | 0.146| -- | -- |\n| Far3D | VoV-99 | (960, 640) | **0.244** | **0.181**| [config](projects/configs/far3d.py) | [model](https://github.com/megvii-research/Far3D/releases/download/v1.0/iter_82548.pth)/[log](https://github.com/megvii-research/Far3D/releases/download/v1.0/far3d.log)|\n\n**Notes**\n- [This config](projects/configs/far3d.py) can be used to reproduce the results on Argoverse 2.\n- For nuScenes version, due to the inconsistent data and evaluation processes, we do not incorporate it to this repo. One can transfer our model part to StreamPETR repo for nuScenes dataset. \n\n## Acknowledgements\n\nWe thank these great works and open-source codebases:\n\n* 3D Detection.  [StreamPETR](https://github.com/exiawsh/StreamPETR), [MMDetection3d](https://github.com/open-mmlab/mmdetection3d), [DETR3D](https://github.com/WangYueFt/detr3d), [PETR](https://github.com/megvii-research/PETR), [BEVFormer](https://github.com/fundamentalvision/BEVFormer), [SOLOFusion](https://github.com/Divadi/SOLOFusion), [Sparse4D](https://github.com/linxuewu/Sparse4D).\n\n\n## Citation\n\nIf you find Far3D is useful in your research or applications, please consider giving us a star 🌟 and citing it by the following BibTeX entry.\n```bibtex\n@article{jiang2023far3d,\n  title={Far3D: Expanding the Horizon for Surround-view 3D Object Detection},\n  author={Jiang, Xiaohui and Li, Shuailin and Liu, Yingfei and Wang, Shihao and Jia, Fan and Wang, Tiancai and Han, Lijin and Zhang, Xiangyu},\n  journal={arXiv preprint arXiv:2308.09616},\n  year={2023}\n}\n```","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fmegvii-research%2FFar3D","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fmegvii-research%2FFar3D","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fmegvii-research%2FFar3D/lists"}