Design and implementation of VGD6-NET framework for waste segregation using 3-tier convolutional neural networks

Waste segregation is an essential process in waste management. It entails identifying, classifying, separating, and arranging various kinds of waste. An efficient waste segregation process facilitates effective reuse, recycling and recovery. Various studies related to waste segregation so far lacking thorough pre-processing approaches might result in diminished classification metrics. Further, several investigations overlooked essential metrics such as F1-score, recall, and precision focusing solely on accuracy. This study proposes novel VGD6-NET architecture (referred as visual garbage detector 6-Net) for 6 categories of waste using 3-tier convolutional neural network. The main aim of this research is to enhance waste image classification by improving accuracy and detection mechanisms through the utilization of advanced technologies such as 3-tier convolutional neural networks and the development of a specialized architecture. The experimental results show the proposed model predicts the categories of waste more accurately for processes like automated waste segregation into recyclable and non-recyclable categories. Trained on 2527 waste images from 6 classes, the model achieved an accuracy score of 0.9854, with a minimal loss score of 0.0814, with a precision of 0.9772 for the cardboard class, recall value of 1.0 for plastic class, and highest F1-score of 0.9764 for the cardboard class of the 6 classes available. Ultimately, the proposed model contributes to building a more sustainable future by reducing the environmental impact of waste disposal and conserving valuable resources through improved recycling efforts.