NMRA-Facilitated Optimized Deep Learning Framework
Abstract
Recycling and landfilling are two of the primary means by which garbage is destroyed in the context of waste management. Many urban areas struggle with improper waste collection, transportation, and disposal. This chapter depicts a competent waste management scheme architecture predicated on internet of things. In addition, two new benchmark datasets to classify waste, which are unified collections of open-source datasets with standardized annotations for all types of waste are presented here. The architecture of the faster region convolutional neural network (FRCNN) is based on the widely used VGG-16 for feature extraction from input images. In addition, the detected garbage is classified into one of seven different types using the naked mole-rat algorithm's (NMRA) hyper-parameter tuning to progress the classification accuracy. The classifier is trained using unlabeled images in a semi-supervised manner. On the test dataset, the proposed method achieves an average precision of 70% in waste detection and an accuracy of 93% in classification.