By taking different crops of an image, a computer network can be trained to recognize features of the entire image, rather than just a specific section. This technique improves image recognition for vision in computer programming.