Real-time Classification of BioGas-Ready and Non-BioGas-Ready Materials in Kitchen Waste using Deep Neural Network

Kitchen waste consists of biodegradable and nonbiodegradable materials. The former produces biogas during decomposition while the latter barely decomposes. Currently, kitchen waste is mostly landfilled in Malaysia. Hence the benefit of biogas from the decomposition of biodegradable waste in not exploited while the toxic elements in the non-biodegradable waste pollutes the soil and underground water. This paper explores the use of deep learning network in classifying these two types of waste. Dataset comprises of real-world kitchen waste images were curated to train YOLOv7 and YOLOv8 for real-time classification. YOLOv8 was found to perform better at an F1-score of 0.649 and 0.776 in recognizing biodegradable and non-biodegradable waste respectively in the test set. Realtime inferencing test results were encouraging at 0.899 and 0.820.