Manu B.N
Thank you so much !!!
Can SAM-HQ support multiple instance segmentation in the same image ? The code and the dataset contain only one mask per image.
Did you try with SD 3.5 ?
I have a 4090 GPU and when I train with BS=1 it gives out of memory error Someone in the other comments section said for BS=1 training one needs minimum...
I made some modifications to gen_crop_data.py to create smaller patches and now I'm able to train on my 4090 GPU