here is what i find like example
https://pytorch.org/tutorials/intermediate/torchvision_tutorial.html
If your training fails with your model and custom dataset, I would recommend to scale down the problem a bit and try to first overfit a small data sample.
E.g. just take 10 samples out of your dataset and try to overfit these samples using your current code.
If that works, you could try to scale up the use case again by using more data.
thank you for the answer acctualy i took juste 90 images and I done a mask for each images but i don’t know if my mask is correct
please can you take a look on what i try to do ?