search

WongKinYiu / CrossStagePartialNetworks

Cross Stage Partial Networks
https://github.com/WongKinYiu/CrossStagePartialNetworks
894 stars 172 forks source link

The figure about yolov4 and yolov5 #36

Closed jzchenriver closed 4 years ago

jzchenriver commented 4 years ago

Hello, I find this figure in the issue, which compares the detection results between different yolo. 84472297-8b0c7c80-acb9-11ea-91a9-49987d70bcfb In the results of each yolov, there are several points. I wanna know the meanings of them. Why does AP increase with average latency increases?

WongKinYiu commented 4 years ago

each point means different input resolution, for example 288, 352, 416, ... and this comparison is out-of-date, it is recommend to find the newest comparison in AlexeyAB's or Ultralytics's repository.

jzchenriver commented 4 years ago

Thank you for your reply. In the paper, you evaluate some detection methods on the mobile GPU (Jetson TX2), and mention that CSPPeleeNet Ref.-EFM (SAM) reaches 41 fps on Jetson TX2. Are these methods included in this repository? Or where I can find it?

WongKinYiu commented 4 years ago

CSPPeleeNet Ref.-EFM (SAM) is a private model. We will release tiny YOLOv4 soon.

jzchenriver commented 4 years ago

I am looking forward to it.

yilawu commented 3 years ago

CSPPeleeNet Ref.-EFM (SAM) is a private model. We will release tiny YOLOv4 soon.

hey, can u release the yolov4-tiny with CSP train code? Thank you.