Instance Segmentation
تجزئة الحالات
Instance segmentation produces a separate pixel mask for every object instance, even if they share the same class. Two adjacent workers receive distinct masks instead of merging into one 'worker' blob. Mask R-CNN was the foundational architecture; YOLO-Seg, Mask2Former, and SAM 2 are the modern fast-and-accurate options. For FI Tech…
Definition
Instance segmentation produces a separate pixel mask for every object instance, even if they share the same class. Two adjacent workers receive distinct masks instead of merging into one 'worker' blob. Mask R-CNN was the foundational architecture; YOLO-Seg, Mask2Former, and SAM 2 are the modern fast-and-accurate options. For FI Tech this matters for counting and tracking — instance segmentation lets us count exactly how many trucks entered a yard today even when six overlap heavily in one frame, and pairs cleanly with multi-object trackers like ByteTrack or BoT-SORT for trajectory reconstruction across cameras.