ivy icon indicating copy to clipboard operation
ivy copied to clipboard

Published 20 hours ago • verified publisher icon nicholaskajoh

yolo.py image_blob = cv2.dnn.blobFromImage

Open jpsolorzanoa opened this issue 2 years ago • 1 comments

Hello, I've setup this repo on wsl2 ubuntu22.04 with opencv 4.5.5 cuda enabled 11.7. Hoever unless I setup this 608 x 608 images size or larger, won't get any detection or will get like ghost detection or frozen boxes. Same video on windows works fine. What could it be?.

image_blob = cv2.dnn.blobFromImage(image, scale, (608, 608), (0, 0, 0), True, crop=False) Screenshot 2022-05-21 221522

jpsolorzanoa avatar May 22 '22 03:05 jpsolorzanoa

Hmmm... Interesting! I have no idea. I've not run Ivy on WSL before. But I don't think the detection and tracking issues are being caused by it (primarily at least). Ivy could use some (actually, lots of) improvement but with what exists, you will get better results by:

  • Using a camera/video with a high frame rate and resolution. Of course this means slower processing so it's a trade-off.
  • Tweaking the MCDF, MCTF and DI to suit your video. See https://github.com/nicholaskajoh/ivy/blob/master/settings.py#L50-L72.

nicholaskajoh avatar May 24 '22 04:05 nicholaskajoh