A Novel Approach for Waste Classification Using Active Deep Learning

Classification of waste on a large scale is complicated, making it difficult to sort and recycle the waste. Many existing systems use models like CNN and architectures such as ResNet, DenseNet, etc. These models perform well in waste classification, but the classification of waste remains complicated due to the inherent variability and ambiguity in waste images. Existing models may struggle with uncertain or ambiguous data, leading to misclassification and reduced overall accuracy. To address these challenges, combining active learning with deep learning can be beneficial. Active deep learning is a concept that improves the model by eliminating uncertainty about the data. This approach will enhance model performance and outperform existing systems. By developing a model for waste classification that combines ResNet50 and active learning, uncertainty and losses can be reduced, resulting in better performance compared to existing systems.