search

doleron / yolov5-opencv-cpp-python

Example of using ultralytics YOLO V5 with OpenCV 4.5.4, C++ and Python
MIT License
329 stars 103 forks source link

8fps on 3070TI #8

Open Pro100rus32 opened 2 years ago

Pro100rus32 commented 2 years ago

I installed opencv with CUDA with version 4.5.4 image I run the file yolo.py image I get 8 - 10 fps, which is too low. I removed the imshow and the fps stayed at the same level. I have noticed that the video card has almost no load. The load is 3 - 6%. I also tried running a smaller model, but fps did not rise above 16. image I ran this on the CPU as well, and got about 4 fps. I tried to run the second script, but it didn't work.

Traceback (most recent call last):
  File "C:\Users\test\Desktop\yolov5-opencv-cpp-python-main\yolo-tiny.py", line 19, in <module>
    input_image = format_yolov5(image) # making the image square
  File "C:\Users\test\Desktop\yolov5-opencv-cpp-python-main\yolo-tiny.py", line 12, in format_yolov5
    row, col, _ = frame.shape
AttributeError: 'NoneType' object has no attribute 'shape'

How can I improve the speed?

freedomcoming commented 2 years ago

Do you use cuda to run the script？ like this: cd yolov5-opencv-cpp-python python python/yolo.py cuda

Pro100rus32 commented 2 years ago

Do you use cuda to run the script？ like this: cd yolov5-opencv-cpp-python python python/yolo.py cuda

Yes. I have 140fps on yolov4 script, but yolov5 working in 10 fps : (

Neo-YH commented 2 years ago

Same issue, CUDA available but low FPS?

fatandfat commented 2 years ago

I found simply change result.setPreferableTarget(dnn::DNN_TARGET_CUDA_FP16); in load_net() to result.setPreferableTarget(dnn::DNN_TARGET_CUDA); will solve the low fps problem

doleron commented 2 years ago

Sorry, bro. I only tested it on note with RTX 2080. I'd keep it out to make easier reach by someone with same issue.