How to extract the images patches from the feature maps after convolution operation?

1

Hi all
I don’t know how to extract the images patches from the feature maps after convolution operation in forward propagation. Because, I want to use this images patches to do some operations.
I know that this operation which extract the images patches is to use the function such as tf.image.extract_image_patches in Tensorflow , but I can’t find the similar function in Pytorch.

So… Could you tell me how to extract the images patches in Pytorch? Thanks！

2

Hello,

You could return the feature maps in forward pass, and use torch.Tensor.unfold to extract the images patches.

3

Thanks for your reply! Your suggestions are really helpful.

4

I am facing the same problem. Were you able to figure out how to do this?