Image Segmentation Basics

Pixel-level understanding

Segmentation classifies each pixel — separating an object's exact shape from the background. It powers medical imaging, self-driving perception, and photo editing.

Two flavors

• Semantic — every pixel gets a class (road, car, sky), but objects of the same class merge.
• Instance — separates individual objects (car 1 vs car 2).

Tooling

Models like U-Net (medical) and the Segment Anything Model (SAM) make segmentation accessible. SAM can segment almost anything from a single click or prompt — a great place to start experimenting.