In computer vision, keypoints are distinctive locations in an image that are likely to be repeatable and identifiable in other images of the same scene or object. They often correspond to corners, blobs, or T-junctions in the image – areas with high local variation in intensity. Keypoint detection algorithms (like Harris, DoG for SIFT, FAST, etc.) find such points automatically. After detecting keypoints, one usually computes descriptors (like SIFT, SURF, ORB descriptors) that characterize the local neighborhood of the keypoint. These descriptors allow matching keypoints between images (for tasks like image stitching, 3D reconstruction, tracking). Keypoints can also be used for object recognition; they act as the “parts” or anchor points for an object’s identity. In deep learning, you might also hear “keypoints” in pose estimation (body joints) or in context of specific models detecting interest points. Generally, keypoints reduce an image to a set of salient points and their descriptors, enabling more efficient and robust matching and analysis.