Resource Constrained Semantic Segmentation for Waste Sorting
This addresses the problem of environmental waste management for Materials Recovery Facilities, though it appears incremental with existing methods applied to new data.
This work tackles efficient waste sorting by developing resource-constrained semantic segmentation models that fit within a 10MB memory limit for edge applications, achieving positive results with minimal impact on Mean IoU through quantization, pruning, and a novel loss function combination.
This work addresses the need for efficient waste sorting strategies in Materials Recovery Facilities to minimize the environmental impact of rising waste. We propose resource-constrained semantic segmentation models for segmenting recyclable waste in industrial settings. Our goal is to develop models that fit within a 10MB memory constraint, suitable for edge applications with limited processing capacity. We perform the experiments on three networks: ICNet, BiSeNet (Xception39 backbone), and ENet. Given the aforementioned limitation, we implement quantization and pruning techniques on the broader nets, achieving positive results while marginally impacting the Mean IoU metric. Furthermore, we propose a combination of Focal and Lovász loss that addresses the implicit class imbalance resulting in better performance compared with the Cross-entropy loss function.