Garbage classification based on fine-tuned state-of-the-art models
Ramil Shukurov · 2023
The world faces one of the crucial issues - waste management, which needs to be tackled urgently to minimize environmental pollution. The main challenge is to classify the massive amount of waste reliably and accurately. Many existing deep learning models employed to achieve accurate results still require to be improved as their performance varies on a different dataset. This research proposes an algorithm based on fine-tuned convolutional neural networks (GoogleNet, ResNet, DenseNet, ResNeXt, EfficientNet) using two distinct optimization algorithms (SGD momentum, Adam) to classify an open-source dataset into twelve classes. Experimental results reveal that the fine-tuned ResNeXt neural network model achieves high model accuracy, 95 %, with a small number of epochs. The experiment indicates that our achievement outperforms several counterpart methods.