Martin Boot Occlusion Image Set

The retail e-commerce industry faces significant challenges in accurately recognizing products that are partially obscured, such as boots with loose laces or those camouflaged by mud. Existing solutions often struggle with this occlusion problem, leading to misidentification and poor customer experience. This dataset aims to address the technical issue of object recognition under occlusion and fulfill the business need for reliable product classification. The dataset consists of images collected from various retail environments, featuring boots in different occlusion scenarios. Data was collected using high-resolution cameras in well-lit conditions, ensuring clarity. Quality control measures include multi-round annotations, consistency checks among labelers, and expert reviews to guarantee high accuracy. The images are stored in JPG format, organized systematically based on occlusion types and visibility metrics.