Instance Segmentation

Instance segmentation combines the capabilities of object detection and semantic segmentation. It not only detects the location of an object and assigns it a class label but also generates a pixel-level mask for every individual object. This allows the model to distinguish between separate instances of the same object class, such as multiple people, vehicles, or animals in a single image.

Many real-world applications require more than simply recognizing object categories. Instance segmentation provides detailed object boundaries and distinguishes individual objects, enabling more accurate scene understanding, object tracking, measurement, and interaction in complex environments.