Extremely Low FPS (1 FPS) When Running YOLO on Intel NUC 7 i5 with Darknet ROS

[Lucas191105-Nanda]

Hello, I hope you are doing well. First of all, thank you for maintaining this project and providing such a great tool for the community. I have been using YOLO with Darknet ROS and successfully integrated it into my system. However, I am facing a significant performance issue when deploying it on my robot, and I would really appreciate any guidance or suggestions to resolve this problem.

I have successfully integrated YOLO with Darknet ROS and can run it without any issues. However, when I deploy it on my robot, which uses an Intel NUC 7 with an i5 processor, I experience extremely low FPS (around 1 FPS).

I have already tried various configuration optimizations, but none of them seem to improve the performance.

I expect to get at least 20 fps when i run the yolo

System (please complete the following information):

• OS: Ubuntu 20.04
• ROS version: ROS Noetic
• GPU when relevant: Intel HD Graphics 620

[1508189250]

I also measured 1ftp around, I took a look at the time mainly in this function：float* prediction = network_predict(net_, X);I feel like I need cuda to speed it up

[Lucas191105-Nanda]

I also measured 1ftp around, I took a look at the time mainly in this function：float* prediction = network_predict(net_, X);I feel like I need cuda to speed it up

what hardware that you use specifically?