Image Segmentation
The task of assigning pixels or regions to semantic or instance-level classes rather than only locating objects with boxes.
Technical Context
Image segmentation assigns labels or instances at pixel or region level, providing spatial detail beyond a bounding box. It is used in forensic image analysis, medical imaging, and scene understanding, where boundary quality and class accuracy should be evaluated separately.
Spatial Boundary
Segmentation produces a denser output than object detection and generally requires more expensive annotation. A good class label does not guarantee an accurate object boundary.
Related Concepts
- Object Detection
- Intersection over Union
- Computer Vision
- Mask
Related article: Image Segmentation