Computer vision is a branch of artificial intelligence that enables computers to interpret and understand images and video in ways that approximate human sight. It combines image processing, machine learning and deep learning to extract meaningful information from visual data.

Modern systems often rely on convolutional neural networks, which learn hierarchical features directly from pixels. Early layers detect simple patterns such as edges and textures, while deeper layers capture complex structures like objects or scenes. Training typically requires large labeled datasets and significant computational power, but once trained, models can perform tasks such as classification, detection and segmentation with high accuracy.

Object detection focuses on identifying and locating multiple items in an image using bounding boxes. Semantic and instance segmentation go further by assigning a label to each pixel, enabling fine grained understanding of scenes. Image recognition and retrieval allow efficient categorization and search within massive image collections.

Beyond visible light, computer vision is applied to infrared, hyperspectral and satellite imagery. In Earth observation, vision models monitor deforestation, urban expansion and crop health, and detect changes in land use over time. In medicine, they assist in analyzing radiological scans, highlighting anomalies that may require attention. In autonomous vehicles and robotics, vision systems support navigation, obstacle avoidance and interaction with dynamic environments.

Current research seeks more robust models that generalize across conditions such as lighting changes, occlusion and viewpoint variation, while reducing data and compute requirements. There is also growing attention to interpretability and trust, aiming to understand why models make particular decisions and to ensure reliable performance in safety critical applications.