What's the difference between classification and segmentation in deep learning?
In particular, can the classification loss function be used for segmentation problems?
Classification generally refers to the problem of classifying an entire entity. For example, in ImageNet, each image is given a class, regardless of whether the relevant information is only a subset of the image.
Segmentation refers to the problem of dividing up and then classifying parts of an image. For example: imagine an image of a cow. Classification would be saying this image contains a cow. Segmentation would try and divide which pixels are cow, which are grass, etc.
Both classification and segmentation are often done with a variety of loss functions. The most popular loss for classification is probably cross entropy, which may be appropriate for some forms of segmentation as well.