Comparison of Garbage Classification Frameworks Using Transfer Learning and CNN

Abstract

With the never-ending increase in the population, garbage and other waste materials have become one of the major hurdles in forming a healthy environment. The proliferation in the development of such schemes and integration of technology brings up the concept of smart waste management based on its biodegradability. These proposed models can be found useful to the smart waste development program and other likely schemes which require the classification of garbage based on their images. The experiment uncovers the reasons behind the working of VGG19 and A9 architecture CNN-based models which were found to provide the best results in accurately detecting the type of garbage. Experimental evaluation was based on 27 models including out of which A9 and VGG19 models were found to be the most efficient ones with 92.24% and 86.35% accuracy, respectively, which are further compared in detail for understanding these models better.