A botanical research app lets users upload a single photo of a plant, and the app must return the most likely species name for the entire image. There is no need to locate or count multiple objects within the picture. Which computer vision technique best fits this requirement?