Is there a function to extract image patches in PyTorch?

Question

Given a batch of images, I'd like to extract all possible image patches, similar to a convolution. In TensorFlow, we can use tf.extract_image_patches to achieve this. Is there an equivalent function in PyTorch?

Thank you.

Answer 1

Unfortunately, there might not be a direct way to achieve your goal.
But Tensor.unfold function might be a solution.
https://discuss.pytorch.org/t/how-to-extract-smaller-image-patches-3d/16837/2
This website might help you.