Definition
An extension of semantic segmentation that also distinguishes between different instances of the same class.
Detailed Explanation
Instance segmentation combines object detection and semantic segmentation to identify individual object instances and their precise boundaries. Popular approaches include Mask R-CNN and other two-stage detectors that first locate objects and then perform pixel-wise segmentation within each detection.
Use Cases
1. Robot manipulation tasks 2. Scene understanding 3. Industrial automation 4. Retail analytics