Down Jacket Vest Occlusion Image Dataset

In the current retail e-commerce industry, the challenge of accurately recognizing clothing items that are partially occluded (e.g., due to overlapping clothing or missing parts) is significant. Existing datasets often do not cover such scenarios adequately, leading to insufficient training for models. This dataset aims to address the technical issue of recognizing and segmenting images of down jacket vests that are obscured by other garments, fulfilling a crucial business need for e-commerce platforms to improve visual search and recommendation systems. The dataset was collected using high-quality cameras in various retail environments, ensuring diverse lighting and angles. To maintain high quality, we implemented multiple rounds of annotation with consistency checks and expert reviews. The images are stored in JPG format, organized by categories based on the occlusion types, facilitating easy access and use.