Waste management is an essential component of urban infrastructure as it maintains proper and efficient disposal of waste generated by human activity. Waste management practices and sorting methods help reduce pollution, preserve natural resources, and uphold public health. A custom-prepared image dataset (1339) consisting of common household waste items (plastic bottle, books, paper, polythene cover, mask, cardboard, newspaper, onion, tomato, carrot, capsicum, brinjal, banana) classified into two categories of waste: dry (725) and wet (589) were used for training pre-trained YOLOv8 models (nano, medium, and large) and obtained the best models. The objective of this approach is to accurately detect and distinguish between dry and wet waste. The obtained model weights were used for a comparative assessment which comprised recall, precision, and F1 score. The medium model resulted in an accuracy 92.02%, precision 89.06%, recall 90.48%, and an F1 score 89.76% for dry waste class, while the wet waste class resulted in an accuracy 81.79%, precision 81.33%, recall 81.67%, and an F1 score 81.50% and performed better when compared with the obtained results of other models.