An Automated Approach to Waste Classification Using Deep Learning

Waste management has been done by humans through direct monitoring and classification of the products or items that are to be sorted-out. Waste management are carried out in industries, hospitals, hotel and tourism, food and beverages, military and other fields to classify the wastes as recyclable and non-recyclable. In this research the waste management is implemented to predict and identify the images into three classes organic, recyclable and non-recyclable using a convolutional neural network model with the inception-net layers. The study developed a custom inception model by adding additional layers and compares the performance through accuracy against the basic Inceptionv3 model. The study used SGD (stochastic gradient descent) with liner regression algorithm for classification and categorical cross-entropy for loss estimation. The current study uses the ReLU function to overcome the under-fitting and over-fitting issues. Dataset was taken from open-source data base. The basic model gained 77% accuracy whereas the custom model obtained 94% accuracy with minimal loss value of 3.