Glossary
Computer vision
A field of AI that trains models to interpret and extract information from images and video, such as detecting objects or faces.
Computer vision is a field of artificial intelligence that trains models to interpret and extract information from images and video, such as identifying objects, faces, text, or activity within a frame.
Modern computer vision systems are almost entirely built on deep learning, specifically convolutional and transformer-based neural network architectures trained on large labeled image datasets, replacing earlier approaches based on hand-engineered features. This differs from audio analytics and speech analytics in the type of raw signal being processed, though multimodal systems increasingly combine vision with audio and text.
Computer vision underlies video analytics applications like counting foot traffic or detecting objects in security footage, as well as content moderation, quality inspection in manufacturing, and remote sensing analysis of satellite imagery. The main pitfalls are dataset bias, a model performing poorly on populations or conditions underrepresented in training data, and overconfidence in edge cases the model was never validated against, which matters especially in safety-critical applications.
Last reviewed September 22, 2026