Question about prediction #17
Description
Let's say I want to use the pre-trained DeepLabV3+ model on an incomplete image of the Cityscapes dataset (i.e. an image with an hole in the middle). Is there a way to do not predict the pixel in the "hole"?
{{ message }}