Deep Learning-Based Construction and Demolition Plastic Resource Recovery

This thesis develops artificial intelligence methods to identify and recover plastic waste from construction and demolition sites. Construction and demolition waste is one of Australia’s largest waste streams, and plastics in this waste are still poorly recovered. These plastics are often mixed, contaminated, fragmented, and difficult to sort manually. The research uses images and computer vision models to recognise plastic types, locate individual plastic objects, and support scalable automated waste auditing. This work supports improved plastic sorting, reduced landfill disposal, and more efficient recycling and circular use of construction materials.