The classification of recycling wastes is of great importance both environmentally and economically. Correct classification of recyclable wastes such as packaging wastes increases the efficiency of the recycling process. This classification process can be done according to the raw material type, colour, shape, size and source of the waste. Correct classification of recycling wastes also provides economic benefits by ensuring more efficient use of resources. The traditional waste classification method involves manually sorting waste into different categories. This method requires a lot of labour and is time consuming. The traditional waste classification method is also prone to human error, which can lead to contamination of recyclable materials. Deep neural networks can quickly identify different types of recyclable materials by analysing images of waste materials. Thus, it can increase efficiency and reduce pollution by sorting them appropriately. In this study, an experimental study was carried out on a data set consisting of 6 classes and 2527 images under the name of "Garbage classification". In this study, a comparative analysis was carried out using the Convolutional Neural Network architectures Resnet101, Convnext and Densenet121. As a result of this study, Resnet101 architecture was more successful than other architectures with an accuracy rate of 98.41%.
Primary Language | English |
---|---|
Subjects | Image Processing, Deep Learning, Artificial Intelligence (Other) |
Journal Section | Research Articles |
Authors | |
Publication Date | December 31, 2023 |
Submission Date | July 31, 2023 |
Published in Issue | Year 2023 Issue: 055 |