Leveraging the Power of Object Detection Models in Identifying Litter for a Significant Reduction in Environmental Pollution

The growing concern of litter pollution in natural environments has escalated into a significant issue that demands immediate and efficient resolution. Recent studies have used deep learning models to solve the problem of litter pollution, but these approaches have faced challenges in accurately detecting litter in real-world environments. Therefore, this paper has proposed a litter detection model and analyze its performance on the TACO dataset, which contains real-world outdoor environment images. The paper evaluates three distinct deep learning models (YOLOv4, YOLOv5, Faster R-CNN) and identifies the best performing model. The performance of the selected model is then enhanced through adjustments of hyperparameters, use of several preprocessing techniques and data augmentation techniques. The experimental results showed that YOLOv5x achieved 88% mAP@.5 and 71.4% mAP@.75 on testing dataset which outperformed the state-of-art studies. The findings of this paper provide valuable insights into the solution of litter pollution and can inform future research in this area.