Improving Plastic Waste Segregation with Deep Learning Techniques

Plastic waste has been infamous for having an upper hand in world-wide pollution levels. While there have been ways to segregate plastic waste from the mixture of wastes, there have been very few solutions tackling the issue of separating the type of plastics, which emphasizes its significance in the recycling process. Segregation of plastics also further leads to better quality of recycled products, conservation of resources such as oil and gas and overall reduction of impact of plastic wastes on the environment. We aim to provide a solution tackling segregation using computer vision deep learning techniques such as RetinaNet and YOLOv8 algorithms to detect and classify types of plastic from a mixture of pre-classified plastic waste. We have been able to perform classification of 6 different types of plastic through our chosen deep learning models. The system's performance was evaluated using measures like F1-score and mean Average Precision (mAP). The efficacy of the strategy, drawbacks, and potential practical use cases were examined. The conclusion lays the groundwork for future research scope in this area while highlighting deep learning's contribution to better plastic waste segregation