Participatory Single-Object Waste Dataset from Mobile Citizen-Science Capture (ScanYourTrash Workshop1, video frames)

This dataset contains annotated video-derived RGB image frames of single post-consumer plastic waste objects collected during the first ScanYourTrash participatory workshop in a school. The workshop was conducted as part of the RecycleBot research project to explore citizen-science–driven data acquisition for computer vision applications in plastic waste sorting. Participants recorded short videos of individual plastic items using their personal smartphones in a controlled indoor classroom setting. Frames were extracted at a rate of one frame per second to balance viewpoint diversity and temporal redundancy. A total of 850 frames were selected and manually annotated with bounding boxes in YOLO format across 21 material-function classes, covering common post-consumer plastics such as PET, HDPE, PP, PS, LDPE, and composite packaging. The dataset follows a 70/15/15 stratified train/validation/test split and is primarily designed for single-object object detection research. It supports methodological studies on participatory data collection, natural versus synthetic augmentation, and class imbalance effects. Due to its controlled acquisition context and predominantly single-object framing, it is not intended as a direct benchmark for industrial sorting deployment without additional domain adaptation. This dataset accompanies and supplements the paper “Scan Your Trash: Exploring Participatory Data Capture to Enrich Object Detection Datasets for Post-Consumer Plastic Sorting,” to be presented at the 33rd CIRP Conference on Life Cycle Engineering (LCE 2026), providing the annotated data and evaluation artifacts underlying the study. For detailed documentation of the data collection protocol, annotation process, dataset structure, evaluation methodology, and limitations, please refer to the accompanying Data Card included in this repository.