Enhanced Waste Segregation Using Vision Transformers and YOLO

Effective waste management is important due to rapid urbanization, meaning cities generate more waste. Managing it properly is important for the environment. The main challenge in the waste management process is separating waste into biodegradable (organic) and non biodegradable (inorganic) categories. Traditional methods of waste segregation are time-consuming, so there is a need for advanced technology. Our proposed state-of-the-art method for real-time object detection, YOLO, simplified the detection process, reduced the computational resources needed and made it faster and more efficient. To make Vision Transformers (ViTs) perform better, we used methods that enhance how they extract features and fit different datasets during training. That includes TACO and TrashNet. We perform hyperparameter tuning to optimize detection accuracy. The process yields better results for materials like clear glass (90%) and PET plastic (85%), while e-waste detection is only challenging for 45% of materials. The proposed method is revolutionizing waste segregation, minimizing labor and contributing to smart city initiatives. Future work will concentrate on enhancing classification accuracy for visually similar materials and e-waste materials.