{"id":13441244,"url":"https://github.com/abhi1kumar/SeaBird","last_synced_at":"2025-03-20T11:38:05.552Z","repository":{"id":230433055,"uuid":"778500294","full_name":"abhi1kumar/SeaBird","owner":"abhi1kumar","description":"[CVPR 2024] Official PyTorch Code of SeaBird: Segmentation in Bird's View with Dice Loss Improves Monocular 3D Detection of Large Objects","archived":false,"fork":false,"pushed_at":"2024-10-27T17:22:51.000Z","size":22097,"stargazers_count":81,"open_issues_count":1,"forks_count":5,"subscribers_count":2,"default_branch":"main","last_synced_at":"2024-10-27T20:37:25.265Z","etag":null,"topics":["3d-computer-vision","3d-object-detection","autonomous-driving","autonomous-vehicle","autonomous-vehicles","bev-segmentation","cvpr","cvpr2024","dice","dice-loss","kitti360","monocular-3d-detection","monocular-3d-localization","nuscenes","object-detection"],"latest_commit_sha":null,"homepage":"https://arxiv.org/pdf/2403.20318.pdf","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/abhi1kumar.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":"2024-03-27T20:42:53.000Z","updated_at":"2024-10-27T17:22:55.000Z","dependencies_parsed_at":"2024-10-27T18:56:57.976Z","dependency_job_id":null,"html_url":"https://github.com/abhi1kumar/SeaBird","commit_stats":null,"previous_names":["abhi1kumar/seabird"],"tags_count":0,"template":false,"template_full_name":null,"repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/abhi1kumar%2FSeaBird","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/abhi1kumar%2FSeaBird/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/abhi1kumar%2FSeaBird/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/abhi1kumar%2FSeaBird/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/abhi1kumar","download_url":"https://codeload.github.com/abhi1kumar/SeaBird/tar.gz/refs/heads/main","host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":221759944,"owners_count":16876323,"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-computer-vision","3d-object-detection","autonomous-driving","autonomous-vehicle","autonomous-vehicles","bev-segmentation","cvpr","cvpr2024","dice","dice-loss","kitti360","monocular-3d-detection","monocular-3d-localization","nuscenes","object-detection"],"created_at":"2024-07-31T03:01:31.561Z","updated_at":"2025-03-20T11:38:05.534Z","avatar_url":"https://github.com/abhi1kumar.png","language":"Python","funding_links":[],"categories":["Python"],"sub_categories":[],"readme":"\u003cdiv align=\"center\"\u003e\n\n\u003cimg src=\"PanopticBEV/images/cvlab.png\" align=\"right\" width=8%\u003e\n\n# SeaBird: Segmentation in Bird's View with Dice Loss Improves Monocular 3D Detection of Large Objects\n\n### [KITTI-360 Demo](https://www.youtube.com/watch?v=SmuRbMbsnZA) | [nuScenes Demo] | [Project](http://cvlab.cse.msu.edu/project-seabird.html) | [Talk](https://www.youtube.com/watch?v=-6CxDymLZ38) | [Slides](https://docs.google.com/presentation/d/15OqUDMIo8vf96vByifpv2npq3aq4xmS38HwjT1wt31U/) | [Poster](https://docs.google.com/presentation/d/15XmI7EAMUzj__5b_Pvdo1c0LTSlqBnONWy3upgfxWfM/)\n\n[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/seabird-segmentation-in-bird-s-view-with-dice/3d-object-detection-from-monocular-images-on-7)](https://paperswithcode.com/sota/3d-object-detection-from-monocular-images-on-7?p=seabird-segmentation-in-bird-s-view-with-dice) \t\n[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/seabird-segmentation-in-bird-s-view-with-dice/3d-object-detection-on-nuscenes-camera-only)](https://paperswithcode.com/sota/3d-object-detection-on-nuscenes-camera-only?p=seabird-segmentation-in-bird-s-view-with-dice)\n\n[![arXiv](http://img.shields.io/badge/arXiv-2403.20318-B31B1B.svg)](https://arxiv.org/abs/2403.20318)\n[![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](https://opensource.org/licenses/MIT)\n[![Visitors](https://api.visitorbadge.io/api/visitors?path=abhi1kumar%2FSeaBird\u0026labelColor=%23FFFFFF\u0026countColor=%23721e82\u0026style=flat)](https://visitorbadge.io/status?path=abhi1kumar%2FSeaBird)\n[![GitHub Stars](https://img.shields.io/github/stars/abhi1kumar/SeaBird?style=social)](https://github.com/abhi1kumar/SeaBird)\n\n\n   \u003cp style=\"font-size:1.2em\"\u003e\n      \u003ca href=\"https://sites.google.com/view/abhinavkumar\"\u003e\u003cstrong\u003eAbhinav Kumar\u003c/strong\u003e\u003c/a\u003e\u003csup\u003e1\u003c/sup\u003e ·\n      \u003ca href=\"https://yuliangguo.github.io\"\u003e\u003cstrong\u003eYuliang Guo\u003c/strong\u003e\u003c/a\u003e\u003csup\u003e2\u003c/sup\u003e ·\n      \u003ca href=\"https://scholar.google.com/citations?user=cL4bNBwAAAAJ\u0026hl=en\"\u003e\u003cstrong\u003eXinyu Huang\u003c/strong\u003e\u003c/a\u003e\u003csup\u003e2\u003c/sup\u003e ·\n      \u003ca href=\"https://www.liu-ren.com\"\u003e\u003cstrong\u003eLiu Ren\u003c/strong\u003e\u003c/a\u003e\u003csup\u003e2\u003c/sup\u003e ·\n      \u003ca href=\"http://www.cse.msu.edu/~liuxm/index2.html\"\u003e\u003cstrong\u003eXiaoming Liu\u003c/strong\u003e\u003c/a\u003e\u003csup\u003e1\u003c/sup\u003e\u003cbr\u003e\n      \u003csup\u003e1\u003c/sup\u003eMichigan State University, \u003csup\u003e2\u003c/sup\u003eBosch Research North America, Bosch Center for AI\n   \u003c/p\u003e\n\n\n\nin [CVPR 2024](https://cvpr.thecvf.com/Conferences/2024/)\n   \u003cp align=\"center\"\u003e\n      \u003cimg src=\"Seabird_teasor.gif\" width=\"784\"\u003e\n   \u003c/p\u003e\n\u003c/div\u003e\n\n\u003e Monocular 3D detectors achieve remarkable performance on cars and smaller objects. However, their performance drops on larger objects, leading to fatal accidents. Some attribute the failures to training data scarcity or the receptive field requirements of large objects. In this paper, we highlight this understudied problem of generalization to large objects. We find that modern frontal detectors struggle to generalize to large objects even on nearly balanced datasets. We argue that the cause of failure is the sensitivity of depth regression losses to noise of larger objects. To bridge this gap, we comprehensively investigate regression and dice losses, examining their robustness under varying error levels and object sizes. We mathematically prove that the dice loss leads to superior noise-robustness and model convergence for large objects compared to regression losses for a simplified case. Leveraging our theoretical insights, we propose SeaBird (Segmentation in Bird's View) as the first step towards generalizing to large objects. SeaBird effectively integrates BEV segmentation on foreground objects for 3D detection, with the segmentation head trained with the dice loss. SeaBird achieves SoTA results on the KITTI-360 leaderboard and improves existing detectors on the nuScenes leaderboard, particularly for large objects.\n   \u003cp align=\"center\"\u003e\n      \u003cimg src=\"PanopticBEV/images/seabird_kitti360_demo.gif\" width=\"784\"\u003e\n   \u003c/p\u003e\n\n\n\n## Citation\n\nIf you find our work useful in your research, please consider starring the repo and citing:\n\n```Bibtex\n@inproceedings{kumar2024seabird,\n   title={{SeaBird: Segmentation in Bird's View with Dice Loss Improves Monocular $3$D Detection of Large Objects}},\n   author={Kumar, Abhinav and Guo, Yuliang and Huang, Xinyu and Ren, Liu and Liu, Xiaoming},\n   booktitle={CVPR},\n   year={2024}\n}\n```\n\n## Single Camera (KITTI-360) Models\n\nSee [PanopticBEV](PanopticBEV)\n\n#### Model Zoo\n\nWe provide logs/models/predictions for the main experiments on KITTI-360 Val /KITTI-360 Test data splits available to download here.\n\n| Data_Splits | Method  | Config\u003cbr/\u003e(Run)                                          | Weight\u003cbr\u003e/Pred  | Metrics | Lrg\u003cbr/\u003e(50) | Car\u003cbr/\u003e(50) | Mean\u003cbr/\u003e(50) | Lrg\u003cbr/\u003e(25) | Car\u003cbr/\u003e(25) | Mean\u003cbr/\u003e(25) | Lrg\u003cbr/\u003eSeg | Car\u003cbr/\u003eSeg | Mean\u003cbr/\u003eSeg | \n|------------|---------|------------------------------------------------------------------|----------|--------|----------|-----------|----------|-----------|----------------|----|--------------|-----------|-----------|\n| KITTI-360 Val  | Stage 1 | [seabird_val_stage1](experiments/pbev_seabird_kitti360_val_stage1.ini) | [gdrive](https://drive.google.com/file/d/17z02Tj92rQtYYVtiUd2O0grkmQ_ZGgdV/view?usp=sharing) | IoU  |   -   |   -   |   -   |   -   |   -   |   -   | 23.83 | 48.54 | 36.18 \n| KITTI-360 Val  | PBEV+SeaBird | [seabird_val](experiments/pbev_seabird_kitti360_val.ini)          | [gdrive](https://drive.google.com/file/d/1iam4F50jX6Hf0WfvkIzkHRkNw_rq3U_p/view?usp=sharing) | AP   | 13.22 | 42.46 | 27.84 | 37.15 | 52.53 | 44.84 | 24.30 | 48.04 | 36.17\n| KITTI-360 Test | PBEV+SeaBird | [seabird_test](experiments/pbev_seabird_kitti360_test.ini)        | [gdrive](https://drive.google.com/file/d/1YC3cWVOmX7bdoU21URnP24KlZLOF4q-D/view?usp=sharing) | AP   |   -   |   -   | 4.64  |   -   |   -   | 37.12 |   -   |   -   |   -   \n\n\n## Multi-Camera (nuScenes) Models\n\nSee [HoP](HoP)\n\n#### Model Zoo\n\n**nuScenes Val Results**\n\n|          Model           | Resolution | Backbone     | Pretrain | APLrg |  mAP   |  NDS   |                      Ckpt/Log/Pred                       |\n| :----------------------: | :--------: | :----------: | :------: | :---: | :----: | :----: | :------------------------------------------------------: |\n| [HoP_BEVDet4D_256](configs/hop_bevdet/hop_bevdet4d-r50-depth.py)    | 256x704  |  ResNet50   | ImageNet-1K | 0.274 | 0.399 | 0.509 | [ckpt](https://github.com/Sense-X/HoP/releases/download/Release/HoP_BEVDet_ep24_ema.pth) / [log](https://github.com/Sense-X/HoP/releases/download/Release/HoP_BEVDet.log) |\n| [HoP+SeaBird_256 Stage1](configs/hop_bevdet/hop_seabird_r50_256x704_stage1.py)    | 256x704  |  ResNet50   | ImageNet-1K | - | - | -| [gdrive](https://drive.google.com/file/d/1zQOO2A4Twno6C1nC53RCoazDVuZ_9W0e/view?usp=sharing) | \n| [HoP+SeaBird_256](configs/hop_bevdet/hop_seabird_r50_256x704.py)    | 256x704  |  ResNet50   | ImageNet-1K | 0.282 | 0.411 | 0.515 | [gdrive](https://drive.google.com/file/d/1dmVQW8yDE423mm6AKBM2o1_3F3V59dXQ/view?usp=sharing) |   \n| [HoP+SeaBird_512 Stage1](configs/hop_bevdet/hop_seabird_r101_512x1408_stage1.py)   | 512x1408 |  ResNet101   | ImageNet-1K | - | - | - | [gdrive](https://drive.google.com/file/d/1D44imXsFSYg9WE-kdnE-SH-G_gjXvxlw/view?usp=sharing) |\n| [HoP+SeaBird_512](configs/hop_bevdet/hop_seabird_r101_512x1408.py)   | 512x1408 |  ResNet101   | ImageNet-1K | 0.329 | 0.462 | 0.547 | [gdrive](https://drive.google.com/file/d/1pgMzLGjXh5A_P3XR7CmQju5qXenTAXAJ/view?usp=sharing) |\n| [HoP+SeaBird_640 Stage1](configs/hop_bevdet/hop_seabird_vov99_640x1600_stage1.py)   | 640x1600 |  V2-99   | DDAD15M | - | - | - | [gdrive](https://drive.google.com/file/d/1cbVkituogo_e5ILMrC8Z8NdemrfQL2uV/view?usp=sharing) |\n| [HoP+SeaBird_640](configs/hop_bevdet/hop_seabird_vov99_640x1600.py)   | 640x1600 |  V2-99   | DDAD15M | 0.403 | 0.527 | 0.602 | [gdrive](https://drive.google.com/file/d/1dz1w0DQrjgw1xm6u6Kp4csNThYkYPFhu/view?usp=sharing) |\n\n**nuScenes Test Results**\n\n|          Model           | Resolution | Backbone     | Pretrain | APLrg |  mAP   |  NDS   |                      Ckpt/Log/Pred                       |\n| :----------------------: | :--------: | :----------: | :------: | :---: | :----: | :----: | :------------------------------------------------------: |\n| [HoP+SeaBird_512 Test](configs/hop_bevdet/hop_seabird_r101_512x1408_test.py)   | 512x1408 |  ResNet101   | ImageNet-1K | 0.366 | 0.486 | 0.570 | [gdrive](https://drive.google.com/file/d/1Y39kVTdw0OXN45u6UedgXfYhePFZ-0Xt/view?usp=sharing) |\n| [HoP+SeaBird_640 Val](configs/hop_bevdet/hop_seabird_vov99_640x1600.py)   | 640x1600 |  V2-99   | DDAD15M | 0.384 | 0.511 | 0.597 | [gdrive](https://drive.google.com/file/d/1dz1w0DQrjgw1xm6u6Kp4csNThYkYPFhu/view?usp=sharing) |\n\n## Acknowledgements\nWe thank the authors of the following awesome codebases:\n- [PanopticBEV](https://github.com/robot-learning-freiburg/PanopticBEV)\n- [BBAVectors](https://github.com/yijingru/BBAVectors-Oriented-Object-Detection) \n- [DEVIANT](https://github.com/abhi1kumar/DEVIANT.git)\n- [DOTA_devkit](https://github.com/CAPTAIN-WHU/DOTA_devkit)\n- [HoP](https://github.com/Sense-X/HoP)\n\nPlease also consider citing them.\n\n## Contributions\nWe welcome contributions to the SeaBird repo. Feel free to raise a pull request.\n\n### \u0026#8627; Stargazers\n[![Stargazers repo roster for @nastyox/Repo-Roster](https://reporoster.com/stars/abhi1kumar/SeaBird)](https://github.com/abhi1kumar/SeaBird/stargazers)\n\n### \u0026#8627; Forkers\n[![Forkers repo roster for @nastyox/Repo-Roster](https://reporoster.com/forks/abhi1kumar/SeaBird)](https://github.com/abhi1kumar/SeaBird/network/members)\n\n\n## License\nSeaBird code is under the [MIT license](https://opensource.org/license/mit).\n\n## Contact\nFor questions, feel free to post here or drop an email to this address- ```abhinav3663@gmail.com```\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fabhi1kumar%2FSeaBird","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fabhi1kumar%2FSeaBird","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fabhi1kumar%2FSeaBird/lists"}