Object Recognition

Object recognition is the cognitive capacity to identify a physical object as an instance of a category — a chair, a face, a violin — despite variation in viewpoint, illumination, occlusion, and context. The problem is formally difficult because the same category can produce infinitely many distinct retinal images, and the same retinal image can be produced by infinitely many distinct three-dimensional configurations. The visual system solves this not by storing templates but by extracting feature hierarchies — from edges and textures in early visual areas to object parts and configurations in higher areas — that are progressively invariant to irrelevant transformations. Object recognition is not a single process but a cascade of inferences, each level betting on the most probable world structure consistent with the level below. The phenomenon connects to broader questions about categorization, concepts, and consciousness: to recognize is already to classify, and to classify is already to impose structure that the world does not independently announce.