Obstructed Images of Casual Shoes

The retail e-commerce industry faces challenges in accurately identifying products in images due to various obstructions like low light conditions and complex foot poses. Existing datasets often lack diversity in occluded scenarios, making it difficult for machine learning models to generalize. This dataset aims to address these issues by providing a comprehensive collection of images featuring casual shoes under obstructed conditions. Data is collected using high-quality cameras in different environments, ensuring variability in lighting and occlusion. Quality control measures include multiple rounds of annotation and expert validation to maintain high accuracy. The dataset is stored in JPEG format, organized by categories of occlusion and lighting conditions. The core advantages of this dataset lie in its high-quality annotations and innovative data collection methods. With an annotation accuracy of over 95% and consistency checks, it significantly outperforms existing datasets. The introduction of advanced image augmentation techniques has improved model performance metrics by up to 30%, addressing real-world e-commerce challenges effectively.