Computer vision has two main tasks: recognition and segmentation. Recognition identifies labels while segmentation breaks down a visual scene into useful entities, and it's important to consider both when building computer vision systems.