A computer-vision demo using Mask R-CNN and YOLOv3 to classify clothing categories and segment individual garments in images — the groundwork behind visual search and outfit-recommendation features.
Approach
Instance segmentation separates each garment as its own region rather than a single bounding box, which matters when multiple articles overlap in a photo. Classification on top of those regions maps each piece to a clothing category.
The demo video walks through the Mask R-CNN / YOLO pipeline on fashion imagery.