Optimizing Waste Classification in Recycling Systems through Transfer Learning and Layer Freezing
R. Rajadevi, R.S. Latha, K. Logeswaran, K. S. Rahshitha, K. Soumiya, P. S. Pavithra · 2025
Waste disposal has always been a critical concern throughout history and remains an important issue today and for the future. Improper waste management can lead to health risks, environmental pollution, and other problems. Addressing this challenge requires an efficient solution, and one such approach involves building an advanced trash identification and classification system using different transfer learning techniques. This problem has been tackled using four different models. The four models are ConvNeXt, EfficientNet B0, ResNet50, and ResNet101. Each of the four models is exposed to distinct layer-freezing techniques that are applied in various ways. With an accuracy of 98.24%, the ConvNeXt model has been determined to be the best model. Therefore, a comparative analysis of the transfer learning models is employed to propose a solution for accurate garbage and recyclable material categorization.