Thanks. Image viewers may interpret the pixels as squares even though they are rectangular, which is why they appear stretched, while video viewers automatically apply the stretch and the video displays correctly. My question is the following: I have this video of dimensions 1440x1080 and extracting the video frames what happens is that I open the image it appears deformed, but I don't know if this is just a display problem or not. What I would like to understand is if it is possible to create a dataset of images directly with the video frames as they are therefore with dimensions 1440x1080 (which appear a little stretched and deformed when opening the image) or is this wrong and must necessarily be resized to 1920x1440?