What Is Computer Vision?

Computer vision is a field of artificial intelligence and electrical engineering concerned with enabling machines to derive meaningful information from images, video, and other visual inputs. It applies signal processing, statistical learning, and geometric reasoning to extract representations of the world, such as the locations of objects, the identities of faces, the three-dimensional structure of a scene, or the actions of people in a video. The field draws on optical physics, applied mathematics, and, increasingly, deep learning, which has displaced many classical approaches since the early 2010s by producing dramatic improvements in recognition accuracy.

Computer vision is distinct from image processing, which applies transformations to image signals such as noise reduction, sharpening, and contrast adjustment, without necessarily interpreting semantic content. Computer vision uses the outputs of image processing as inputs to higher-level inference. The field overlaps with machine perception, robotics, and medical image analysis, and its research is published in venues including the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), one of the highest-impact conferences in the computing research community.

Image Understanding and Object Recognition

Object recognition is the task of identifying what objects are present in an image and, in the case of object detection, where they are located. Classification assigns a category label to an entire image. Detection additionally produces bounding boxes around individual instances. Semantic segmentation assigns a category label to every pixel. These tasks are hierarchically related: segmentation requires detection, which requires recognition. Convolutional neural networks (CNNs) substantially advanced accuracy on all three tasks following the 2012 ImageNet competition, where AlexNet reduced the classification error rate by a large margin compared to prior methods. Subsequent architectures including VGG, ResNet, and Vision Transformers have continued the progression. Face recognition, a specialization of object recognition, raises distinct concerns about privacy and civil liberties that are the subject of active policy debate alongside the technical research.

Pose Estimation and Activity Recognition

Human pose estimation infers the positions of body joints from images or video, producing a skeletal representation of how a person is positioned or moving. Single-person and multi-person pose estimation methods differ substantially in computational approach because the multi-person case requires simultaneously detecting individuals and associating keypoints across bodies. Activity recognition builds on pose estimation and temporal modeling to classify actions or behaviors across video sequences. A person walking, raising their hand, or assembling an object are distinguishable activities when temporal context is incorporated into the classifier. These capabilities depend on large annotated datasets; the COCO keypoints dataset and Kinetics video dataset are among the benchmarks used to measure progress. Applications in elder care monitoring, physical rehabilitation, and sports analytics have driven sustained research interest.

Deep Learning Foundations and Architectures

The transformation of computer vision by deep learning rests on the ability of convolutional and attention-based architectures to learn hierarchical feature representations directly from labeled examples. Earlier approaches required manual feature engineering, such as SIFT and HOG descriptors, which captured specific aspects of texture, edge, or gradient patterns but were designed by hand. Deep networks learn representations at multiple spatial scales simultaneously, from low-level edges to high-level semantic concepts. Transfer learning, in which a network pre-trained on a large dataset such as ImageNet is fine-tuned on a smaller domain-specific dataset, has made high-quality vision models accessible even when labeled training data is limited. The IEEE Transactions on Pattern Analysis and Machine Intelligence is a primary journal publishing foundational research on visual learning and recognition algorithms.

Applications

Computer vision has applications in a wide range of fields, including:

• Autonomous vehicles, using cameras and depth sensors to detect obstacles, lanes, and pedestrians
• Medical image analysis, detecting abnormalities in radiology scans, pathology slides, and retinal images
• Indoor navigation and robotics, mapping environments and localizing mobile platforms
• Industrial quality inspection, identifying surface defects and assembly errors on production lines
• Security and surveillance, detecting unauthorized access and tracking movement in monitored spaces