Application of deep learning-based approaches in classifying environmental wastes

The waste problem has emerged worldwide due to distorted urbanization, industrialization, and people's unconscious actions. This problem negatively affects our lives and necessitates living in an environment with pollution. Every piece of randomly discarded waste, whether intentional or unintentional, becomes part of nature. As a result, the self-disposal of these wastes in nature takes years. The environmental pollution problem is minimized by separating the waste. Thus, contributing to recycling extends the average lifespan by reducing environmental pollution. In this study, the benefit of recycling was mentioned by using Deep Learning models and machine learning methods. The classification process of environmental waste was carried out using the waste separation dataset and architectures such as AlexNet, GoogLeNet, ResNet50, DenseNet201, ShuffleNet, SqueezeNet. In the next stage, feature maps obtained from deep architectures were extracted. Results were also obtained from the classification methods of support vector machines. As a result of the experiments obtained from the proposed hybrid model, it was observed that the best accuracy was 85.96%.