SortWaste: A Densely Annotated Dataset for Object Detection in Industrial Waste Sorting

TL;DR

The paper tackles the challenge of automated waste sorting in real-world industrial settings by introducing SortWaste, a densely annotated dataset collected from a Material Recovery Facility, and ClutterScore, a frame-level metric quantifying visual complexity. It benchmarks multiple state-of-the-art object detectors (e.g., Faster R-CNN, TridentNet, RetinaNet, YOLOv11) on full and plastic-only subsets, and analyzes performance as a function of scene clutter. The results show promising plastic-detection performance (up to $mAP=0.597$ for YOLOv11 on plastic, with $AP_{50}$ around $0.75$), but reveal a clear degradation in highly cluttered scenes, underscoring the need for clutter-robust models. By making SortWaste publicly available and providing ClutterScore, the work lays a foundation for developing practical, industrially relevant detection systems that can operate under real-world clutter and contamination conditions.

Abstract

The increasing production of waste, driven by population growth, has created challenges in managing and recycling materials effectively. Manual waste sorting is a common practice; however, it remains inefficient for handling large-scale waste streams and presents health risks for workers. On the other hand, existing automated sorting approaches still struggle with the high variability, clutter, and visual complexity of real-world waste streams. The lack of real-world datasets for waste sorting is a major reason automated systems for this problem are underdeveloped. Accordingly, we introduce SortWaste, a densely annotated object detection dataset collected from a Material Recovery Facility. Additionally, we contribute to standardizing waste detection in sorting lines by proposing ClutterScore, an objective metric that gauges the scene's hardness level using a set of proxies that affect visual complexity (e.g., object count, class and size entropy, and spatial overlap). In addition to these contributions, we provide an extensive benchmark of state-of-the-art object detection models, detailing their results with respect to the hardness level assessed by the proposed metric. Despite achieving promising results (mAP of 59.7% in the plastic-only detection task), performance significantly decreases in highly cluttered scenes. This highlights the need for novel and more challenging datasets on the topic.

Figures (7)

• Figure 1: Sample images from the proposed waste dataset, illustrating diverse waste categories with bounding-box annotations for object detection.
• Figure 2: Representative samples from several state-of-the-art waste detection datasets. Each image illustrates a typical example from a different dataset, with the dataset titles displayed above the respective samples.
• Figure 3: Illustration of the data collection setup, showing the arrangement of waste on the conveyor belt, the flow direction, the position of the workers, and the image capture system using a smartphone on a tripod.
• Figure 4: Representative examples from the SortWaste dataset, illustrating different categories of waste items on a conveyor belt.
• Figure 5: Distribution of annotation areas across waste classes. Boxplots show the variation in object sizes (in pixels²) for each category.