Waste Classification Using Improved Deep Learning Method
Jianchun Qi, Minh Nguyen, Wei Qi Yan · Advances in computational intelligence and robotics book series · 2025
In this book chapter, YOLOv8, the effective version in the YOLO series, is modified through data augmentation, context strategies, and an advanced attention mechanism. These modifications aim to primarily improve the quality of waste dataset and the classification accuracy of small objects within given waste classes, thereby boosting the overall performance of the model. The waste data is classified into four classes, and 1,000 waste images were labelled for model training. Upon evaluation, the classification accuracy of the improved model reached 85.6%. The effectiveness of these improvements was further substantiated through ablation studies.