Developing household solid waste management models for waste valorization in Dhaka City by integrating artificial intelligence

The management of the organic fraction of household solid waste (HSW) in Dhaka city poses significant challenges due to the lack of a proper HSW collection system, and the lack of integration of automated technologies and artificial intelligence (AI) into the waste segregation and sorting stage. Therefore, this study aims to introduce two innovative HSW management (HSWM) models for the valorization of the wastes. The first model features a double-compartment bin that facilitates the source separation of the organic and inorganic HSWs at homes using greenish and yellowish biodegradable bags. In this model, an automated system that uses camera technology in conjunction with an optical reader will detect the colored packed waste bags when the bags move along a conveyor. The second model proposes a single-compartment bin where all types of HSW (i.e., mixed wastes) are collectively packed in one biodegradable bag. This model incorporates a developed MobileNetV2 model of AI to classify the organic and inorganic HSWs, which will lessen the waste segregation and sorting time and enhance the effectiveness of the model. The MobileNetV2 model is developed using a transfer learning model of AI that incorporates deep learning techniques with sophisticated convolutional neural networks (CNN) for improving the classification accuracy. The MobileNetV2 demonstrates training, testing, and validation accuracy of 94.86%, 89.08%, and 87.11% respectively. The MobileNetV2 successfully classifies dates fruit, eggshell, assorted sweet, and sweet potato as organic with confidence levels 99.9%, 99.9%, 88.8%, and 96.9 % respectively, while also identifying metallic water bottle, plastic pen, styrofoam, and drinking glass as inorganic with confidence levels 98.8%, 87.4%, 99.6%, and 100.0% respectively. The confusion matrix of MobileNetV2 reveals the true positives and true negatives values of 177 and 276, respectively. Additionally, the precision and recall values for organic items are 85% and 88%, respectively, while for inorganic items, these are 92% and 90%. The proposed HSWM models will guide all the involved stakeholders in the HSWM chain to valorize the HSW in Dhaka city and contribute to enhancing the circular economy (CE) practices in Bangladesh.