Akademik Veri Yönetim Sistemi
Signal, Image and Video Processing, cilt.19, sa.8, 2025 (SCI-Expanded)
Recycling plays a vital role in conserving natural resources, minimizing environmental pollution, saving energy, and promoting sustainable living. Addressing the issue of recycling through humans, who are the source of waste generation, is considered the cheapest and most effective method. In this study, a smart recycling bin with an embedded system has been designed to both raise environmental awareness among children and more effectively teach them to use recycling bins correctly, even without supervision. There are recycling bin studies with deep learning for waste classification; however, there is no study specifically designed as a recycling bin to teach children the correct classification of waste, as in this study. This smart recycle bin has an image processing system controlled by a Raspberry Pi 4B processor. When a waste is placed in any compartment of the recycle bin, images of the waste are obtained and classified with a trained deep neural network. It is determined if the waste is placed into the correct compartment. If the waste is placed in the wrong compartment, the cover remains closed, and the child is guided to the correct compartment through entertaining content on the LCD screen of the recycling bin. A pre-trained MultiNet neural network is used for the classification of waste images. The MultiNet model consists of a combination of the CNN pre-trained models DenseNet-201, NASNetMobile and VGG16. A dataset containing 1075 images of paper, plastic, glass, and metal is created. The classification accuracy 98% is obtained using 7-fold cross-validation.