Abstract
With the aim to tackle the issue of waste classification for different categories of misspend substances, the authors, with a limited availability of dataset have processed a highly accurate model to classify garbage into 7 different categories using the CompostNet dataset. Experiments were carried out on pre-trained models of MobileNetV2, ResNet34 and Densenet121 model, previously trained on ImageNet dataset. The accuracies obtained were 96.42%, 96.27% and 96.273% respectively for the Densenet121, mobilenetv2 and resnet34 models. Within 60 epochs, the neural network model accurately categorizes waste materials provided in the input image. The results of the experiments are compared with other previous work done in the same field. The applications of the experiments conducted in this research aims at providing better waste categorization and also follows the United Nations goal for Responsible Consumption and Production towards sustainable development.
Highlights
With the huge amount of garbage produced by consumers every day all over the world, the waste materials are mixed with each other without proper categorization and segregation
Comparisons of the results with the ones provided by Umut Ozkaya & Levent Seyfi [8] are mentioned which show high accuracy on the Trashnet dataset with fine-tuned Convolutional Neural Network [9] models through SoftMax classifiers
With the boom of Deep Learning in the recent years, image classifiers have been studied in the field of Convolutional Neural Networks (CNNs) [9]
Summary
With the huge amount of garbage produced by consumers every day all over the world, the waste materials are mixed with each other without proper categorization and segregation. 90% of this is collected out of which, only 1/5th of it stands processable and the remaining waste goes to the dump sites with all sorts of harmful e-wastes, plastic, liquid waste, partially consumed food, unfinished water bottles, variety of metals, poly-ethene bags, etc. This diversified accumulation of waste materials makes it extremely hard for useful materials such as bio-degradable and recycle-able items to be categorized and gathered to be utilized. With the advancement of deep learning and technology in specific, humans tend to utilize them for the good and for a social cause for the betterment of society and eventually integrate the same with the goals of sustainable development and less exploitation of materials on earth
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
