[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"project-71174":3},{"id":4,"name":5,"fullName":6,"owner":7,"repo":5,"description":8,"homepage":9,"htmlUrl":10,"language":11,"languages":10,"totalLinesOfCode":10,"stars":12,"forks":13,"watchers":14,"openIssues":15,"contributorsCount":16,"subscribersCount":16,"size":16,"stars1d":17,"stars7d":18,"stars30d":19,"stars90d":16,"forks30d":16,"starsTrendScore":20,"compositeScore":21,"rankGlobal":10,"rankLanguage":10,"license":22,"archived":23,"fork":23,"defaultBranch":24,"hasWiki":25,"hasPages":23,"topics":26,"createdAt":10,"pushedAt":10,"updatedAt":29,"readmeContent":30,"aiSummary":31,"trendingCount":16,"starSnapshotCount":16,"syncStatus":32,"lastSyncTime":33,"discoverSource":34},71174,"RT-DETR","lyuwenyu\u002FRT-DETR","lyuwenyu","[CVPR 2024] Official RT-DETR (RTDETR paddle pytorch), Real-Time DEtection TRansformer, DETRs Beat YOLOs on Real-time Object Detection. 🔥 🔥 🔥 ","",null,"Python",5282,621,36,411,0,14,30,94,42,108.78,"Apache License 2.0",false,"main",true,[27,28],"rtdetr","rtdetrv2","2026-06-12 04:00:59","English | [简体中文](README_cn.md)\n\n\n\u003Ch2 align=\"center\">RT-DETR: DETRs Beat YOLOs on Real-time Object Detection\u003C\u002Fh2>\n\u003Cp align=\"center\">\n    \u003C!-- \u003Ca href=\"https:\u002F\u002Fgithub.com\u002Flyuwenyu\u002FRT-DETR\u002Fblob\u002Fmain\u002FLICENSE\">\n        \u003Cimg alt=\"license\" src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FLICENSE-Apache%202.0-blue\">\n    \u003C\u002Fa> -->\n    \u003Ca href=\"https:\u002F\u002Fgithub.com\u002Flyuwenyu\u002FRT-DETR\u002Fblob\u002Fmain\u002FLICENSE\">\n        \u003Cimg alt=\"license\" src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Flicense\u002Flyuwenyu\u002FRT-DETR\">\n    \u003C\u002Fa>\n    \u003Ca href=\"https:\u002F\u002Fgithub.com\u002Flyuwenyu\u002FRT-DETR\u002Fpulls\">\n        \u003Cimg alt=\"prs\" src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fissues-pr\u002Flyuwenyu\u002FRT-DETR\">\n    \u003C\u002Fa>\n    \u003Ca href=\"https:\u002F\u002Fgithub.com\u002Flyuwenyu\u002FRT-DETR\u002Fissues\">\n        \u003Cimg alt=\"issues\" src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fissues\u002Flyuwenyu\u002FRT-DETR?color=pink\">\n    \u003C\u002Fa>\n    \u003Ca href=\"https:\u002F\u002Fgithub.com\u002Flyuwenyu\u002FRT-DETR\">\n        \u003Cimg alt=\"issues\" src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Flyuwenyu\u002FRT-DETR\">\n    \u003C\u002Fa>\n    \u003Ca href=\"https:\u002F\u002Farxiv.org\u002Fabs\u002F2304.08069\">\n        \u003Cimg alt=\"arXiv\" src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FarXiv-2304.08069-red\">\n    \u003C\u002Fa>\n    \u003Ca href=\"mailto: lyuwenyu@foxmail.com\">\n        \u003Cimg alt=\"emal\" src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fcontact_me-email-yellow\">\n    \u003C\u002Fa>\n\u003C\u002Fp>\n\n---\n\n\nThis is the official implementation of papers \n- [DETRs Beat YOLOs on Real-time Object Detection](https:\u002F\u002Farxiv.org\u002Fabs\u002F2304.08069)\n- [RT-DETRv2: Improved Baseline with Bag-of-Freebies for Real-Time Detection Transformer](https:\u002F\u002Farxiv.org\u002Fabs\u002F2407.17140)\n\n\n\u003Cdetails>\n\u003Csummary>Fig\u003C\u002Fsummary>\n\n\u003Ctable>\u003Ctr>\n\u003Ctd>\u003Cimg src=https:\u002F\u002Fgithub.com\u002Flyuwenyu\u002FRT-DETR\u002Fassets\u002F77494834\u002F0ede1dc1-a854-43b6-9986-cf9090f11a61 border=0 width=500>\u003C\u002Ftd>\n\u003Ctd>\u003Cimg src=https:\u002F\u002Fgithub.com\u002Fuser-attachments\u002Fassets\u002F437877e9-1d4f-4d30-85e8-aafacfa0ec56 border=0 width=500>\u003C\u002Ftd>\n\u003C\u002Ftr>\u003C\u002Ftable>\n\u003C\u002Fdetails>\n\n\n## 🚀 Updates\n- \\[2025.11.18\\] Release the **newest** member of the RT-DETR family: [RT-DETRv4:Painlessly Furthering Real-Time Object Detection with Vision Foundation Models](https:\u002F\u002Fgithub.com\u002FRT-DETRs\u002FRT-DETRv4).\nBy harnessing the rapidly evolving capabilities of Vision Foundation Models (VFMs), we boost lightweight detectors and, without incurring any extra inference latency, significantly improve the performance of the full-size model.\n- \\[2024.11.28\\] Add torch tool for parameters and flops statistics. see [run_profile.py](.\u002Frtdetrv2_pytorch\u002Ftools\u002Frun_profile.py)\n- \\[2024.10.10\\] Add sliced inference support for small object detecion. [#468](https:\u002F\u002Fgithub.com\u002Flyuwenyu\u002FRT-DETR\u002Fpull\u002F468)\n- \\[2024.09.23\\] Add ✅[Regnet and DLA34](https:\u002F\u002Fgithub.com\u002Flyuwenyu\u002FRT-DETR\u002Ftree\u002Fmain\u002Frtdetr_pytorch) for RTDETR.\n- \\[2024.08.27\\] Add hubconf.py file to support torch hub.\n- \\[2024.08.22\\] Improve the performance of ✅ [RT-DETRv2-S](.\u002Frtdetrv2_pytorch\u002F) to 48.1 mAP (\u003Cfont color=green>+1.6\u003C\u002Ffont> compared to RT-DETR-R18).\n- \\[2024.07.24\\] Release ✅ [RT-DETRv2](.\u002Frtdetrv2_pytorch\u002F)!\n- \\[2024.02.27\\] Our work has been accepted to CVPR 2024!\n- \\[2024.01.23\\] Fix difference on data augmentation with paper in rtdetr_pytorch [#84](https:\u002F\u002Fgithub.com\u002Flyuwenyu\u002FRT-DETR\u002Fcommit\u002F5dc64138e439247b4e707dd6cebfe19d8d77f5b1).\n- \\[2023.11.07\\] Add pytorch ✅ *rtdetr_r34vd* for requests [#107](https:\u002F\u002Fgithub.com\u002Flyuwenyu\u002FRT-DETR\u002Fissues\u002F107), [#114](https:\u002F\u002Fgithub.com\u002Flyuwenyu\u002FRT-DETR\u002Fissues\u002F114).\n- \\[2023.11.05\\] Upgrade the logic of `remap_mscoco_category` to facilitate training of custom datasets, see detils in [*Train custom data*](.\u002Frtdetr_pytorch\u002F) part. [#81](https:\u002F\u002Fgithub.com\u002Flyuwenyu\u002FRT-DETR\u002Fcommit\u002F95fc522fd7cf26c64ffd2ad0c622c392d29a9ebf).\n- \\[2023.10.23\\] Add [*discussion for deployments*](https:\u002F\u002Fgithub.com\u002Flyuwenyu\u002FRT-DETR\u002Fissues\u002F95), supported onnxruntime, TensorRT, openVINO.\n- \\[2023.10.12\\] Add tuning code for pytorch version, now you can tuning rtdetr based on pretrained weights.\n- \\[2023.09.19\\] Upload ✅ [*pytorch weights*](https:\u002F\u002Fgithub.com\u002Flyuwenyu\u002FRT-DETR\u002Fissues\u002F42) convert from paddle version.\n- \\[2023.08.24] Release RT-DETR-R18 pretrained models on objects365. *49.2 mAP* and *217 FPS*.\n- \\[2023.08.22\\] Upload ✅ [*rtdetr_pytorch*](.\u002Frtdetr_pytorch\u002F) source code. Please enjoy it!\n- \\[2023.08.15\\] Release RT-DETR-R101 pretrained models on objects365. *56.2 mAP* and *74 FPS*.\n- \\[2023.07.30\\] Release RT-DETR-R50 pretrained models on objects365. *55.3 mAP* and *108 FPS*.\n- \\[2023.07.28\\] Fix some bugs, and add some comments. [1](https:\u002F\u002Fgithub.com\u002Flyuwenyu\u002FRT-DETR\u002Fpull\u002F14), [2](https:\u002F\u002Fgithub.com\u002Flyuwenyu\u002FRT-DETR\u002Fcommit\u002F3b5cbcf8ae3b907e6b8bb65498a6be7c6736eabc).\n- \\[2023.07.13\\] Upload ✅ [*training logs on coco*](https:\u002F\u002Fgithub.com\u002Flyuwenyu\u002FRT-DETR\u002Fissues\u002F8).\n- \\[2023.05.17\\] Release RT-DETR-R18, RT-DETR-R34, RT-DETR-R50-m（example for scaled).\n- \\[2023.04.17\\] Release RT-DETR-R50, RT-DETR-R101, RT-DETR-L, RT-DETR-X.\n\n## 📣 News\n- RTDETR and RTDETRv2 are now available in Hugging Face Transformers. [#413](https:\u002F\u002Fgithub.com\u002Flyuwenyu\u002FRT-DETR\u002Fissues\u002F413), [#549](https:\u002F\u002Fgithub.com\u002Flyuwenyu\u002FRT-DETR\u002Fissues\u002F549)\n- RTDETR is now available in [ultralytics\u002Fultralytics](https:\u002F\u002Fdocs.ultralytics.com\u002Fzh\u002Fmodels\u002Frtdetr\u002F).\n\n## 📍 Implementations\n- 🔥 RT-DETRv2\n  - paddle: [code&weight](.\u002Frtdetrv2_paddle\u002F)\n  - pytorch: [code&weight](.\u002Frtdetrv2_pytorch\u002F)\n- 🔥 RT-DETR \n  - paddle: [code&weight](.\u002Frtdetr_paddle)\n  - pytorch: [code&weight](.\u002Frtdetr_pytorch)\n\n\n| Model | Input shape | Dataset | $AP^{val}$ | $AP^{val}_{50}$| Params(M) | FLOPs(G) | T4 TensorRT FP16(FPS)\n|:---:|:---:| :---:|:---:|:---:|:---:|:---:|:---:|\n| RT-DETR-R18 | 640 | COCO | 46.5 | 63.8 | 20 | 60 | 217 |\n| RT-DETR-R34 | 640 | COCO | 48.9 | 66.8 | 31 | 92 | 161 |\n| RT-DETR-R50-m | 640 | COCO | 51.3 | 69.6 | 36 | 100 | 145 |\n| RT-DETR-R50 |  640 | COCO | 53.1 | 71.3 | 42 | 136 | 108 |\n| RT-DETR-R101 | 640 | COCO | 54.3 | 72.7 | 76 | 259 | 74 |\n| RT-DETR-HGNetv2-L | 640 | COCO | 53.0 | 71.6 | 32 | 110 | 114 |\n| RT-DETR-HGNetv2-X | 640 | COCO | 54.8 | 73.1 | 67 | 234 | 74 |\n| RT-DETR-R18 | 640 | COCO + Objects365 | **49.2** | **66.6** | 20 | 60 | **217** |\n| RT-DETR-R50 | 640 | COCO + Objects365 | **55.3** | **73.4** | 42 | 136 | **108** |\n| RT-DETR-R101 | 640 | COCO + Objects365 | **56.2** | **74.6** | 76 | 259 | **74** |\n**RT-DETRv2-S** | 640 | COCO  | **48.1** \u003Cfont color=green>(+1.6)\u003C\u002Ffont> | **65.1** | 20 | 60 | 217 |\n**RT-DETRv2-M**\u003Csup>*\u003Csup> | 640 | COCO  | **49.9** \u003Cfont color=green>(+1.0)\u003C\u002Ffont> | **67.5** | 31 | 92 | 161 |\n**RT-DETRv2-M** | 640 | COCO | **51.9** \u003Cfont color=green>(+0.6)\u003C\u002Ffont> | **69.9** | 36 | 100 | 145 |\n**RT-DETRv2-L** | 640 | COCO | **53.4** \u003Cfont color=green>(+0.3)\u003C\u002Ffont> | **71.6** | 42 | 136 | 108 |\n**RT-DETRv2-X** | 640 | COCO | 54.3 | **72.8** \u003Cfont color=green>(+0.1)\u003C\u002Ffont>  | 76 | 259| 74 |\n\n**Notes:**\n- `COCO + Objects365` in the table means finetuned model on COCO using pretrained weights trained on Objects365.\n\n\n## 🦄 Performance\n\n### 🏕️ Complex Scenarios\n\u003Cdiv align=\"center\">\n  \u003Cimg src=\"https:\u002F\u002Fgithub.com\u002Flyuwenyu\u002FRT-DETR\u002Fassets\u002F77494834\u002F52743892-68c8-4e53-b782-9f89221739e4\" width=500 >\n\u003C\u002Fdiv>\n\n### 🌋 Difficult Conditions\n\u003Cdiv align=\"center\">\n  \u003Cimg src=\"https:\u002F\u002Fgithub.com\u002Flyuwenyu\u002FRT-DETR\u002Fassets\u002F77494834\u002F213cf795-6da6-4261-8549-11947292d3cb\" width=500 >\n\u003C\u002Fdiv>\n\n## Citation\nIf you use `RT-DETR` or `RTDETRv2` in your work, please use the following BibTeX entries:\n```\n@misc{lv2023detrs,\n      title={DETRs Beat YOLOs on Real-time Object Detection},\n      author={Yian Zhao and Wenyu Lv and Shangliang Xu and Jinman Wei and Guanzhong Wang and Qingqing Dang and Yi Liu and Jie Chen},\n      year={2023},\n      eprint={2304.08069},\n      archivePrefix={arXiv},\n      primaryClass={cs.CV}\n}\n\n@misc{lv2024rtdetrv2improvedbaselinebagoffreebies,\n      title={RT-DETRv2: Improved Baseline with Bag-of-Freebies for Real-Time Detection Transformer}, \n      author={Wenyu Lv and Yian Zhao and Qinyao Chang and Kui Huang and Guanzhong Wang and Yi Liu},\n      year={2024},\n      eprint={2407.17140},\n      archivePrefix={arXiv},\n      primaryClass={cs.CV},\n      url={https:\u002F\u002Farxiv.org\u002Fabs\u002F2407.17140}, \n}\n```\n","RT-DETR 是一个实时目标检测项目，旨在通过改进的检测变换器（DETR）实现优于YOLO系列模型的实时性能。该项目基于Python开发，支持PaddlePaddle和PyTorch框架，其核心功能包括高效的实时目标检测、对小物体检测的支持以及通过视觉基础模型提升轻量级检测器性能而不增加推理延迟。此外，RT-DETR还提供了多种骨干网络选择如RegNet和DLA34，并且持续更新以优化模型性能。它适用于需要高效准确地进行视频监控、自动驾驶等场景下的实时目标识别任务。",2,"2026-06-11 03:36:25","high_star"]