Functional Beverage Occluded Image Dataset

The current retail e-commerce sector faces challenges with accurately identifying products due to occlusion in images. Existing datasets often lack sufficient examples of occluded products, leading to poor performance in real-world applications. This dataset aims to address the need for robust training data that includes various occlusion levels for functional beverages like Red Bull and Monster. The data was collected using high-resolution cameras in well-lit environments, ensuring clarity even with occlusions present. Quality control measures were implemented, including multiple rounds of labeling and expert reviews to ensure data consistency and accuracy. The dataset is organized in a directory structure with JPG images and accompanying metadata files for easy access and processing. The core advantages of this dataset include high-quality labels with over 95% accuracy, which significantly enhances model performance. Innovative annotation techniques were applied to capture varying occlusion levels, allowing for better generalization in real-world scenarios. This dataset is expected to reduce error rates in product recognition tasks by at least 30% compared to previous datasets, providing substantial improvements in retail inventory management and automated checkout systems.