Recycle AId: Training an Artificial Intelligence Model Based on Computer Vision for Solid Waste Classification

By 2030 and 2050, global solid waste generation is expected to reach 2.01 and 3.40 billion tons per year, respectively. Despite the continuous increase in solid waste production and growing environmental awareness, inadequate education on waste reduction and proper handling practices remains a significant challenge. Improper waste classification contributes to environmental pollution through hazardous waste infiltration into the soil and leads to resource losses when recyclable materials are improperly discarded. To address these challenges, this work proposes a solid waste classification model based on computer vision techniques. The model is designed to classify waste materials into four categories: cardboard, metal, paper, and plastic. Training was performed using a hybrid dataset that combines the publicly available TrashNet dataset with custom images captured under real-world conditions in a specialized enclosure that simulates operational environments. Variants of the ResNet architecture, specifically ResNet-50 and ResNet-152, were fine-tuned using transfer learning to leverage pre-trained image classification layers and optimize performance for the classification task. Experimental results showthat the proposed model achieves precision values of 87.1%, 70.2%, 76.3%, and 70.2% for metal, paper, plastic, and cardboard, respectively.