Short answers, pulled from the story.
Computer vision is an interdisciplinary field concerned with enabling computers to extract high-level understanding from digital images or videos. It produces numerical or symbolic information from visual data, including decisions, three-dimensional models, and object classifications.
Computer vision began in the late 1960s at universities pioneering artificial intelligence. In 1966, researchers believed the problem could be solved in a single undergraduate summer project by attaching a camera to a computer and having it describe what it saw.
Performance of convolutional neural networks on the ImageNet Large Scale Visual Recognition Challenge now approaches human levels. However, these systems still struggle with small or thin objects and filter-distorted images, while they surpass humans at fine-grained classification tasks such as identifying specific breeds or species.
For 2024, the leading markets for computer vision were industry with a market size of 5.22 billion US dollars, medicine at 2.6 billion, and military at 996.2 million.
Medical computer vision extracts diagnostic information from imaging data to detect conditions such as tumors, arteriosclerosis, and dental pathologies, and to measure organ dimensions and blood flow. Progress in convolutional neural networks has improved disease detection specifically in cardiology, pathology, dermatology, and radiology.
Neurobiology directly shaped computer vision algorithms. The Neocognitron, a neural network built by Kunihiko Fukushima in the 1970s, was explicitly modeled on the primary visual cortex. Deep learning methods for image classification also trace their conceptual origins to the study of biological neural systems.