search

AlexeyAB / darknet

YOLOv4 / Scaled-YOLOv4 / YOLO - Neural Networks for Object Detection (Windows and Linux version of Darknet )
http://pjreddie.com/darknet/
Other
21.75k stars 7.96k forks source link

Different GPUs performance #5970

Open windscapemax opened 4 years ago

windscapemax commented 4 years ago

Hi all,

I have a task to analyze multiple video streams from IP cameras. So i am using YoloV3-Tiny to do it. Recently i used Quadro P2000 for multiple instance object detection. I run a few processes (7-8) and it works fine(~12 fps each). Then i want to switch for more powerful GPU and bought GTX1660TI. It looks much powerfull then Quadro P2000, but in practice it lose to P2000. I can run only 2-3 streams. OS: Windows 10 1903 Enterprise Latest drivers, CUDA 10.0, cuDNN also compatible version for CUDA 10.0.

Dose any one know the solution?

Thank you.

KyloEntro commented 4 years ago

How do you use darknet yolo model in your soft ? C++ ? Python ? OpenCV dnn backend ?

windscapemax commented 4 years ago

C++ ? Python ? I am using it with much more primitive way. This project is currently in the concept stage so we are using C#. OpenCV dnn backend ? Yes, i am using EmguCV 4.0.1 version, as i know its wrapper for OpenCV.

KyloEntro commented 4 years ago

Maybe try with opencv C++ first and compare fps