Noe Acache discusses the use of Dino v2 for computing image representations, highlighting its ability to segment important features in pictures using attention maps, while also addressing challenges with matching large distortions in images of similar objects.