Skip to content

Commit

Permalink
Update and rename README_CN.md to README.md
Browse files Browse the repository at this point in the history
  • Loading branch information
LSH9832 authored Dec 7, 2024
1 parent 114c5c6 commit 5351945
Showing 1 changed file with 9 additions and 2 deletions.
11 changes: 9 additions & 2 deletions README_CN.md → README.md
Original file line number Diff line number Diff line change
Expand Up @@ -3,17 +3,24 @@

<div align="center">

[English](README.md)
[English](README_EN.md)

[![arXiv](https://img.shields.io/badge/arXiv-2302.07483-<COLOR>.svg)](https://arxiv.org/abs/2302.07483)

</div>

#### 工具推荐: 使用SAM([Segment Anything Model](https://github.com/facebookresearch/segment-anything)) 辅助的图像标注工具: [SAMLabeler Pro](https://github.com/LSH9832/SAMLabelerPro),支持多人同时远程标注

## 简介
- EdgeYOLO 在嵌入式设备 Nvidia Jetson AGX Xavier 上达到了34FPS,在COCO2017数据集上有**50.6**% AP的准确度,在VisDrone2019-DET数据集上有**25.9**% AP的准确度 **(图像输入大小为640x640, 批大小为16, 包含后处理时间)**。更小的模型EdgeYOLO-S在COCO2017数据集上以**44.1**% AP、**63.3**% AP<sup>0.5</sup>(目前单阶段P5小模型中最好的)准度达到了53FPS的速度。
- 我们提供了更加强大的数据增强方法,可以在数据集标签稀疏时起到更好的效果。
- 在训练末尾阶段使用RH损失函数,中小模型的检测效果有所提升。
- 我们的论文(预印版)已在[**arxiv**](https://arxiv.org/abs/2302.07483)公布

## 更新
**[2024/12/7] 重大更新** <br>
1. 上传了多平台统一化部署项目,无需针对不同平台修改代码,详见目录[deployment](./deployment)

**[2024/3/16]** <br>
1. 上传了[demo/amct_onnx2om.py](https://github.com/LSH9832/edgeyolo/blob/main/demo/amct_onnx2om.py) 该脚本用于将onnx模型转换为华为昇腾系列设备上支持的om模型(比如昇腾310),并且发布[c++部署代码示例](https://github.com/LSH9832/edgeyolo/blob/main/cpp/ascend). (注意,你必须拥有华为官网提供的相应依赖库和工具,部分依赖仅购买相应硬件的用户才有下载权限)

Expand Down Expand Up @@ -324,7 +331,7 @@ make -j4
year={2023}
}
```

[![Star History Chart](https://api.star-history.com/svg?repos=LSH9832/edgeyolo&type=Timeline)](https://star-history.com/#LSH9832/edgeyolo&Timeline)

## 目前发现的bugs
- 在训练时有时可能会触发以下错误,降低pytorch版本至1.8.0应该可以解决这个问题。
Expand Down

0 comments on commit 5351945

Please sign in to comment.