Fine-Grained Attribute Annotation for Fashionpedia
Managed the end-to-end data preparation lifecycle for a specialized computer vision project involving over 48,000 images. I designed and executed a multi-layered labeling strategy across 27 apparel categories and 19 apparel parts, implementing 294 fine-grained attribute labels. My work focused on establishing complex parent-child relationships between labels to support an experimental Mask R-CNN branch for visual attribute detection.