On construction sites, effective waste sorting is critical for both environmental compliance and material recovery value. However, sorting practices are often inconsistent, resulting in contaminated containers that cannot be sold for full recycling value. This leads to higher disposal costs and diminished sustainability impact. Besix, a global construction leader, sought a way to automatically monitor container content and detect sorting errors early, without adding operational overhead.
Besix partnered with Sagacify to develop a solution capable of analyzing construction site imagery and assessing waste sorting quality. Working with limited training data due to pandemic-related site closures, the teams implemented a custom annotation environment and collaborated closely to label real-world images captured on-site.
Sagacify built an AI model that detects containers in construction site images, classifies their contents, and identifies the presence and proportion of contaminants. The system automatically alerts site managers when contamination is detected, enabling faster corrective action. To address the challenge of minimal labeled data, a self-supervised learning approach was used to maximize model performance under low-data conditions. The solution was designed to work with standard crane-mounted cameras and integrates into Besix’s on-site workflows.
Feeling lost ?
Let's discuss over a coffee; our AI experts have you covered!
