Can't use GPU

[DavidWZX]

GPU 2060 CUDA 10.1 Cudnn 7.65 tensorflow 2.3.0rc0

But the GPU didn't be uesd at all when running this code, and I only got about 15 FPS.

Is there any way to solve it?

[jmudy]

Try installing tensorflow-gpu (instead of tensorflow) with pip install tensorflow-gpu==2.3.0rc0 and check if the GPU is used.

[DavidWZX]

Try installing tensorflow-gpu (instead of tensorflow) with pip install tensorflow-gpu==2.3.0rc0 and check if the GPU is used.

Thank you for your advice!

But I have checked my environment again and I did use tensorflow-gpu 2.3.0rc0, you can see that in following picture.

So what about the FPS you got while you are running this code?

[jmudy]

Sorry, I did not see in your previous message the FPS you get. Don't worry, your 15 FPS is great. I checked this repository on Google Colab using an NVIDIA Tesla T4 and got around 14-19 FPS with the video test (yolov4-deepsort/data/video/test.mp4). GPU usage is very high when you run Deepsort with YOLOv4.

[DavidWZX]

Sorry, I did not see in your previous message the FPS you get.

Don't worry, your 15 FPS is great. I checked this repository on Google Colab using an NVIDIA Tesla T4 and got around 14-19 FPS with the video test (yolov4-deepsort/data/video/test.mp4). GPU usage is very high when you run Deepsort with YOLOv4.

Thanks for checking and replying information to me! That's very kind of you.

I checked Task Manager and it only showed less than 10% usage of GPU while running code. But by using command line my GPU have more than 50% usage and nearly all the GPU memory are being occupied. So maybe the Task Manager can't show the real-time GPU state at least on my machine.

Thanks again! ^_^