Remondis Contamination Dataset (RCD)

Le résumé fourni par la source

The Remondis Contamination Dataset (RCD) is a novel and challenging collection of images designed to facilitate the development and evaluation of computer vision models for the detection of plastic-bag contamination in waste collection scenarios. This dataset was meticulously curated from historical records of Remondis, a waste management company, and open source online resources which provides real-world visuals captured by cameras installed on waste collection trucks. The images cover various scenarios including waste bins, truck hoppers, different camera zooms, angles, and lighting conditions, accurately reflecting the complexities of waste collection environments. Dataset Details: Format: JPEG images, RGB color scheme Dimensions: 640 x 480 pixels Resolution: 72 pixels-per-inch Annotations: Bounding box annotations for plastic-bag candidates Annotation Format: KITTI Contaminant Types: Plastic bags including coles bags, woolie bags, color bags, white bags, black bags, and packaging material Total Images: 1125 samples (968 for training, 157 for validation) Total Bounding Box Annotations: 1851 (1588 for training, 263 for validation) Key Challenges: The RCD dataset poses unique challenges due to visual similarities between plastic bags and non-contaminants, such as white paper or dark portions in the image. The diverse range of plastic bag types, along with the intricacies of waste collection environments, make the task of plastic-bag contamination detection particularly demanding. Applications: The RCD dataset serves as a benchmark for practical waste segregation tasks, including the detection of various waste contaminants, characterization of waste contents, and counting occurrences of specific waste types. The dataset is a valuable resource for researchers, practitioners, and enthusiasts working on computer vision applications in waste management and environmental monitoring. Citation: If you use the RCD dataset in your research, please consider citing it using the following reference: U. Iqbal, J. Barthelemy, P. Perez, and T. Davies, “Edge-Computing Video Analytics Solution for Automated Plastic-Bag Contamination Detection: A Case from Remondis,” Sensors, vol. 22, no. 20, p. 7821, Oct. 2022, doi: 10.3390/s22207821.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.