-
Notifications
You must be signed in to change notification settings - Fork 111
Description
Thanks for you github.
I did have to scratch my head to figure out why it wasn't working for me.
1 - In Hailo8I custom object, you do a "cd model" just before the yolo detect which is wrong. (scratch that line)
Now back to the optimize problems.
1 - I'm doing 5000 images with 9 objects.
The optimize.py script needs 46GB to create the array.
If you have not enough memory you should explained that you need to create a swapfile.
I did have to create a 100GB swapfile because of my 5000 images.
2 - The optimize.py works only with RGB images . I did have some Palette ( 8 bits) Jpeg images which stop with errors the optimize.py script.
It could be good to specify that only colors 24 bits images works. (8 bits image works on yolov)
3 - I got that error which show me that input was null. I spend a lot of times to figure it out until I change the classes to 9 in
~/Hailo8l/config/postprocess_config/my_yolov8n_ms_config.json.
Maybe add small note in your description about those problems.
I assume that yolov11 will have the same problem.
After these changes it is working!
regards,
Daniel Perron