MMCDWaste: An multi-label mixed construction and demolition waste dataset

Dataset Description:This paper introduces the first multi-label mixed construction and demolition waste dataset specifically designed for complex real-world recycling scenarios, named MM-CDWaste (Multi-label Mixed Construction and Demolition Waste). The dataset comprises 4,786 high-resolution images, covering nine representative categories of commonly encountered CDW materials: tile, metal, foam, timber, glass, plastic, rubber, brick, and concrete, as depicted in Figure X(a). Unlike existing CDW datasets that primarily focus on single-material cases, MM-CDWaste is uniquely constructed to reflect the heterogeneous mixture and coexistence of multiple waste types in actual demolition environments. Each image contains multiple categories simultaneously, with varying degrees of aggregation and mixing, from mildly scattered distributions to severely blended accumulations, as illustrated in Figure X(b). This design better captures the intrinsic multi-label properties of CDW and the challenges faced in automated sorting systems. Compared with publicly available datasets in related research, MM-CDWaste offers greater category diversity, more realistic material combinations, and a richer representation of cluttered environments, which provides a valuable benchmark for advancing multi-label recognition and intelligent recycling technologies in CDW processing. Due to publication confidentiality requirements, the dataset is temporarily restricted, and only a partial subset is released in this study. The complete dataset will be made publicly available after the acceptance of this article. Usage Policy:Researchers interested in using MM-CDWaste for academic purposes are encouraged to contact us in advance. Proper citation of this dataset is mandatory, and cooperative authorship may be considered depending on the level of contribution and collaboration.