A Deep Learning-Based approach to Segregate Solid Waste Generated in Residential Areas

Abstract

Residential waste is a substantial contributor to solid waste generation, which is approximately around 36.5 million tons annually in India. The waste created in households is not separated at the source. All waste is accumulated in a single waste bin and stashed in a nearby public waste bin, resulting in a massive amount of waste being dumped in landfills and also infused with other types of waste, causing environmental pollution. The core objective of this research is to develop a household waste segregator using the TensorFlow object detection model and Arduino microcontroller. The SSD MobileNet V2 model has been trained with a household dataset consisting of paper, plastic, metal, organic waste, glass, and one more additional empty class to detect whether waste is placed for detection or not. This proposed system can predict the waste class and segregate it into their specific dustbin with mean Average Precision (mAP) and recall of 86.5% and 88.3%, respectively. Waste segregation and recycling can reduce landfills, lower carbon footprints, increase recycling, recover value from garbage, and lower greenhouse gases emitted from waste. Segregation at the source will reduce the cost of the segregation process carried by the municipal solid waste management.