Pytorch transform issue ,

how i can get the orignal image back?

The usage of reshape is wrong as it will interleave the image.
Use permute to permute the dimensions of the tensor instead.

If you want to get the original pixel values back, you would have to “denormalize” the image such that the values are again in the original range (most likely [0, 255]) instead of [0, 1], which is done by ToTensor().

1 Like