← Computer Vision learning pathComputer Vision · Module 6 of 10
Image segmentation
Classify every pixel to outline objects precisely.
📖 Key ideas
- Semantic segmentation labels each pixel by class (road, car, sky).
- Instance segmentation separates individual objects.
- Panoptic combines both.
- Models: U-Net, Mask R-CNN, SAM (Segment Anything).
- Used in medical imaging and autonomous driving.
💡 Real-life example: Segmentation is colouring a colouring book where every region gets exactly the right colour.
Semantic segmentationClass label for every pixel
Instance segmentationSeparate mask per object
MaskPixel-level object outline
U-NetEncoder-decoder segmentation network