Garbage Classification using Deep Learning: A Neural Network Approach for Automated Waste Sorting

Effective waste management has become important for the world to protect the nurture from pollution. The automatic detection and classify of the garbage is essential for managing in separation of waste, control the increasing volume and complexity of household waste. This research focused on deep learning-based image classification system to automatically categorize waste into six classes: cupboard, paper, trash, metal, glass, and plastic. We evaluated and compared the performance of four Neural networks architectures: a custom-designed CNN, MobileNetV2, DenseNet121, and EfficientNetB0. All models trained using the Adam optimizer and categorical cross entropy loss. Among the tested architectures, EfficientNetB0 achieved the validation accuracy of 94.38% and a test accuracy of 94.66%, with the lowest validation loss of 0.1851. On the other hand, the MobileNetV2 gained the accuracy of 92.43% and custom CNN and DenseNet121 models yielded promising results with validation accuracies of 87.95% and 88.96%, respectively. Overall, the performance results demonstrated that deep learning models, particularly EfficientNetB0, could be effectively employed in automated garbage classification systems to support and more scalable waste sorting solutions.

Ҳали таржима қилинмаган