Computer Vision

Short definition

Computer vision is the field of AI focused on visual understanding. It allows systems to detect objects, classify images, read text from photos, track movement, estimate depth and analyze video streams.

How it works

Modern computer vision models often use deep neural networks trained on large image or video datasets. They learn visual features such as edges, textures, shapes and objects, then apply those patterns to new images.

Example

In manufacturing, a vision system can inspect products on a line and flag defects that are too subtle or too fast for reliable manual review. In retail, it can help analyze shelf availability or checkout flows.

Why it matters

Computer vision connects AI to the physical world. It powers medical imaging, robotics, autonomous vehicles, quality control and image search. The risks include privacy concerns, biased datasets and false positives in high-stakes environments.

Common task types

Classification labels an entire image, detection locates objects, segmentation assigns labels to pixels and OCR reads text. The chosen task changes the required annotations, evaluation metrics and deployment cost. A classifier will not automatically show where a defect is located unless it was designed and trained for localization.

Real-world conditions

Performance can fall after a change in camera, perspective, lighting, background or season. Testing should include imperfect conditions and partially hidden objects. Systems that analyze people also need clear retention rules, access controls and a privacy-impact assessment rather than relying only on model accuracy.