A media company wants to automatically generate a short natural-language sentence describing the overall content of each uploaded photograph (for example, 'A group of people walking on a beach at sunset'). This description will be displayed as alt text for accessibility. Which computer vision capability should they use?