How can I use BCELoss with images.
My predictions are of shape (Batch_size,channels,H,W)
Are you working on a binary classification task or a multi-label classification?
The output of your model and the target should have the same shape, so your posted shape might work for a multi-label classification use case.