Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

add rtdetrv2 #9073

Merged
merged 5 commits into from
Jul 30, 2024
Merged

add rtdetrv2 #9073

merged 5 commits into from
Jul 30, 2024

Conversation

MINGtoMING
Copy link
Contributor

add rtdetrv2

Copy link

paddle-bot bot commented Jul 27, 2024

Thanks for your contribution!

# Conflicts:
#	configs/rtdetrv2/README.md
#	configs/rtdetrv2/_base_/rtdetrv2_reader.yml
primaryClass={cs.CV}
}

@software{Lv_rtdetr_by_cvperception_2023,
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

这引用更新下 把software这个换成 v2的文章+

@misc{lv2024rtdetrv2improvedbaselinebagoffreebies,
      title={RT-DETRv2: Improved Baseline with Bag-of-Freebies for Real-Time Detection Transformer}, 
      author={Wenyu Lv and Yian Zhao and Qinyao Chang and Kui Huang and Guanzhong Wang and Yi Liu},
      year={2024},
      eprint={2407.17140},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2407.17140}, 
}

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

好的

<details>
<summary>3. 转换成TensorRT(可选) </summary>

- 确保TensorRT的版本>=8.5.1
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

离散采样的模型 是不需要这个版本要求的;可以再准确描述下

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

已改正

@MINGtoMING MINGtoMING force-pushed the rtdetrv2-dev branch 2 times, most recently from d54f657 to 423aee8 Compare July 29, 2024 13:36
RT-DETRv2是基于 Transformer 的实时端到端检测器。它在SOTA的 RT-DETR
的基础上,引入了灵活的解码器,并运用了一系列有效的训练策略。具体而言,我们为解码器的各种特征图建议了不同数量的采样点,在多个训练阶段采用动态数据增强策略,并为每个独特的模型确定特定的优化超参数。为适应各种部署方案,解码器现在提供了一个利用离散采样而非网格采样的选项。RT-DETRv2-R18
在相同速度下相比 RT-DETR-R18 实现了 1.4 的提升,在 T4 GPU 上以 FP16 模式达到了 47.9 mAP 和 217
FPS。而且,混合精度训练策略的使用使得训练速度提高了 15%,GPU 内存使用减少了 20%
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

训练速度和显存 这块描述也去掉吧

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

ok

Copy link
Collaborator

@lyuwenyu lyuwenyu left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

lgtm

@lyuwenyu lyuwenyu merged commit 3a9f3de into PaddlePaddle:develop Jul 30, 2024
3 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

Successfully merging this pull request may close these issues.

3 participants