Waste Segregation System

Abstract

Effective waste management is crucial for environmental sustainability and automated waste classification systems plays a vital role in this process. The project focuses on developing a waste segregation system utilizing a Raspberry Pi microcontroller along with the VGG-16 Convolutional Neural Network (CNN) model for accurately categorizing waste into paper, plastic, and metal classes. The system utilizes the VGG-16 model for its high accuracy in distinguishing between different waste materials. The system begins by IR sensor detecting the presence of waste and capturing the image of waste item using a Pi Camera. The image is then sent to the trained VGG-16 model which categorizes the waste into one of three classes. The highest prediction category is selected and microcontroller sends a signal to the servo motor which rotates the bin according to the provided signal eliminating the need for manual sorting. For the validation dataset, the model achieved an accuracy of 85.76%, and for the test dataset, the accuracy was 85.44%. The precision, recall, and F1-score metrics demonstrate strong performance across all three waste categories, with paper showing the highest recall at 0.93 in validation and 0.94 in testing. The system successfully segregated waste materials based on their classification into the appropriate bins.