Construction and Demolition Waste Object Detection Dataset (CODD)

The CODD dataset is a meticulously curated collection of images and annotations designed to facilitate the development and benchmarking of both bounding box and instance segmentation detection models for Construction and Demolition Waste (CDW) sorting. CODD features ten distinct CDW categories, including bricks, concrete, tiles, wood, pipes, plastics, general waste, foaming insulation, stones, and plaster boards. The dataset was carefully acquired from a recycling facility in Cyprus, capturing the diverse characteristics of CDW in their natural state. A total of 3,129 high-resolution images (1920 × 1200 × 3, RGB) containing 16,545 annotated samples make up this comprehensive resource. The dataset is divided into training, validation, and testing subsets, with the option for users to exercise discretion in the use of the validation set within their specific research framework. All annotations are provided in the standardized PASCAL VOC XML format, including both bounding box coordinates and polygon coordinates for precise object segmentation detection.