ConMetal-6 subset

Published: 13 June 2026| Version 1 | DOI: 10.17632/ch34f38kz5.1
Contributors:
Diyana Ranasinghe, Mehrdad Arashpour

Description

ConMetal-6 is a fine-grained, alloy-level object detection dataset designed for automated sorting of metal streams in construction and demolition waste (CDW). This publicly released subset constitutes the test split (10%) of the full ConMetal-6 dataset, comprising 232 RGB images and 4,620 bounding-box annotations across six metal alloy classes: Aluminium, Brass, Copper, Ferrous Metal, Galvanized Steel, and Stainless Steel. Images were captured under in-the-wild conditions at varying illumination levels in Melbourne, Australia, using mobile RGB sensors. Ground truth labels were established through magnetic testing, scratch testing, and expert review. Annotations are provided in YOLO format (.txt per image). The complete dataset (2,304 images; 35,787 instances) is available upon request for research purposes.

Files

Institutions

Categories

Computer Vision, Waste Management, Construction Waste

Licence