Classification of TrashNet Dataset Based on Deep Learning Models


Aral R. A., Keskin S. R., Kaya M., Haciomeroglu M.

IEEE International Conference on Big Data (Big Data), Washington, United States Of America, 10 - 13 December 2018, pp.2058-2062 identifier

  • Publication Type: Conference Paper / Full Text
  • Volume:
  • City: Washington
  • Country: United States Of America
  • Page Numbers: pp.2058-2062
  • Keywords: Deep learning, CNN, image classification, waste, recycling, NETWORKS
  • Gazi University Affiliated: Yes

Abstract

Waste recycling is very important in terms of economy and climate balance of the world. For this reason, classifying recyclable garbage is an important goal for humanity and Deep Learning models can be used for this purpose. In this study, we tested well-known Deep Learning models to provide the most efficient approach. In this study, Densenet121, DenseNet169, InceptionResnetV2, MobileNet, Xception architectures were used for Trashnet dataset and Adam and Adadelta were used as the optimizer in neural network models. Based on the findings obtained in this study, Adam provided better test accuracies compared to Adadelta. Besides, the data augmentation process was applied to increase classification accuracy because of limited samples of the Trashnet dataset. As a result of the conducted experiments, the best results were found in the DenseNet121 using fine-tuning with a test accuracy rate of 95%. A similar success rate was also found in the InceptionResNetV2 model using fine-tuning with a test accuracy of 94%.