Yolov5 for Oriented Object Detection

The code for the implementation of “Yolov5 + Circular Smooth Label”.

Results and Models

The results on DOTAv1.5_subsize1024_gap200_rate1.0 test-dev set are shown in the table below. (password:yolo)

Model (link)	Size (pixels)	TTA (multi-scale/ rotate testing)	OBB mAPtest 0.5 DOTAv1.5	Speed CPU b1 (ms)	Speed 2080Ti b1 (ms)	Speed 2080Ti b16 (ms)	params (M)	FLOPs @1024 (B)
yolov5m [baidu/google]	1024	×	73.19	328.2	-	-	21.6	50.5
yolov5m6	1024	×	-	-	-	-	-	-
yolov5m7	1024	×	-	-	-	-	-	-

Table Notes (click to expand)

• All checkpoints are trained to 300 epochs with COCO pre-trained checkpoints, default settings and hyperparameters.
• mAP^test values are for single-model single-scale on DOTAv1.5 dataset.
  Reproduce by python val.py --data 'data/dotav15_poly.yaml' --img 1024 --conf 0.01 --iou 0.4 --task 'test' --batch 16
• Speed averaged over DOTAv1.5 val_split_subsize1024_gap200 images using a 2080Ti gpu. NMS + pre-process times is included.
  Reproduce by python val.py --data 'data/dotav15_poly.yaml' --img 1024 --task speed --batch 1

Updates

• [2022/1/7] : Faster and stronger, some bugs fixed

Installation

Please refer to install.md for installation and dataset preparation.

Getting Started

This repo is based on yolov5. Please see GetStart.md for the Oriented Detection basic usage.

Acknowledgements

I have used utility functions from other wonderful open-source projects. Espeicially thank the authors of:

• ultralytics/yolov5.
• Thinklab-SJTU/CSL_RetinaNet_Tensorflow.
• jbwang1997/OBBDetection
• CAPTAIN-WHU/DOTA_devkit

More detailed explanation

想要了解相关实现的细节和原理可以看我的知乎文章:

• 自己改建YOLOv5旋转目标的踩坑记录.

有问题反馈

在使用中有任何问题，欢迎反馈给我，可以用以下联系方式跟我交流

• 知乎（@略略略）
• 代码问题提issues,其他问题请知乎上联系

关于作者

  Name  : "胡凯旋"
  describe myself："咸鱼一枚"