Optimized and automated methods for handling construction and demolition waste (CDW) are crucial for improving the resource recovery process in waste management. Automated waste recognition is a critical step in this process, and it relies on robust image segmentation techniques. Prompt-guided segmentation methods provide promising results for specific user needs in image recognition. However, the current state-of-the-art segmentation methods trained for generic images perform unsatisfactorily on CDW recognition tasks, indicating a domain gap. To address this gap, a user-guided segmentation pipeline is developed in this study that leverages prompts such as bounding boxes, points, and text to segment CDW in cluttered environments. The adopted approach achieves a class-wise performance of around 70 % in several waste categories, surpassing the state-of-the-art algorithms by 9 % on average. This method allows users to create accurate segmentations by drawing a bounding box, clicking, or providing a text prompt, minimizing the time spent on detailed annotations. Integrating this human-machine system as a user-friendly interface into material recovery facilities enhances the monitoring and processing of waste, leading to better resource recovery outcomes in waste management.
The construction industry generates a substantial volume of solid waste, often destinated for landfills, causing significant environmental pollution. Waste recycling is decisive in managing waste yet challenging due to labor-intensive sorting processes and the diverse forms of waste. Deep learning (DL) models have made remarkable strides in automating domestic waste recognition and sorting. However, the application of DL models to recognize the waste derived from construction, renovation, and demolition (CRD) activities remains limited due to the context-specific studies conducted in previous research. This paper aims to realistically capture the complexity of waste streams in the CRD context. The study encompasses collecting and annotating CRD waste images in real-world, uncontrolled environments. It then evaluates the performance of state-of-the-art DL models for automatically recognizing CRD waste in-the-wild. Several pre-trained networks are utilized to perform effectual feature extraction and transfer learning during DL model training. The results demonstrated that DL models, whether integrated with larger or lightweight backbone networks can recognize the composition of CRD waste streams in-the-wild which is useful for automated waste sorting. The outcome of the study emphasized the applicability of DL models in recognizing and sorting solid waste across various industrial domains, thereby contributing to resource recovery and encouraging environmental management efforts.