WongKinYiu / yolov9

Implementation of paper - YOLOv9: Learning What You Want to Learn Using Programmable Gradient Information
GNU General Public License v3.0
8.97k stars 1.42k forks source link

How about the fps / latency of yolov9 vs other yolo-series network? #23

Closed Yuxinyi-Qiyu closed 8 months ago

Yuxinyi-Qiyu commented 8 months ago

I see the tables in the paper don't have the speed.

WongKinYiu commented 8 months ago

Some of other works use A100 or TRT to produce fps / latency, and currently A100 is unavailable for our team. Maybe a good solution is to install all of these works in a same environment and test their performance. For TRT, in our yolov7 experiments, custom optimized code is about 250% faster than direct converted code. Some works put their efforts on optimized TRT inference code, so it is really hard to make fair comparison using TRT inference.

Yuxinyi-Qiyu commented 8 months ago

ronment and test their performance. For TRT, in our yolov7 experiments, custom optimized code is about 250% faster than direct converted code. Some works put their efforts on optimized TRT inference code, so it is really hard to make fair comparison using TRT inference.

Thank you for your reply!☺️