A review of computer vision applications in litter and cleanliness monitoring

Municipal solid waste management struggles with manual processes, affecting data accuracy and street cleanliness monitoring. Recent research highlights computer vision as a solution for automated litter detection, improving efficiency and reducing costs. This study reviews 65 studies on computer vision in urban waste management, using PRISMA 2020 to address litter and cleanliness in urban areas. The study is divided into three parts: (i) dataset curation, (ii) model training, and (iii) Comparative analysis and challenges. There are five steps in dataset curation: (i) Set the objective, (ii) Acquisition, (iii) Preprocessing, (iv) Annotation, and (v) Splitting. The datasets utilized in these studies range from 114 to 1,10,988 images, encompassing diverse environmental conditions to support the training of machine learning models. Furthermore, the choice of machine learning algorithms employed in these studies is diverse, from traditional methods such as RF (Random Forest) to advanced deep learning techniques like CNN (Convolutional Neural Networks), R-CNN (Region- convolutional neural network), and the recent YOLO (You Only Look Once) model. The studies underscore the extensive application of the F-score metric, alongside other metrics such as accuracy, average precision, error rate, and mean average precision (mAP), with F-score values reported to reach as high as 0.93.