Automatic Trash Can with Artificial Intelligence Technology

Abstract

This research aims to develop an automated waste bin capable of waste classification using automation and artificial intelligence (AI) technology to enhance waste management efficiency. The system employs an ESP32-CAM as the primary device to capture waste images and transmit data to an AI model developed using a convolutional neural network (CNN) on the Teachable Machine platform. This model classifies waste into four categories: general waste, hazardous waste, organic waste, and recyclable waste. Upon classification, the ESP32-CAM transmits the results to an Arduino Uno R3, which controls the waste sorting mechanism using motors and servo motors. Additionally, an infrared (IR) sensor is installed to monitor the waste bin's status and trigger an alert when the bin is full. The system's performance was evaluated by classifying 50 waste samples per category and assessing four key criteria: (1) classification accuracy, (2) accuracy of the waste container movement, (3) ability to halt operation when the waste bin is full, and (4) LED notification functionality. Experimental results showed that the system achieved a maximum classification accuracy of 99.56% for organic waste due to its distinct characteristics, while general waste had the lowest accuracy of 96.00% due to similarities with certain recyclable waste types. Furthermore, the system demonstrated an accuracy of over 90% in directing the waste container to the correct bin across all categories, with a 100% accuracy in the hazardous waste category. The system also consistently performed stably in alerting when the waste bin was full using the infrared sensor. Overall, the test results indicate that the developed automated waste bin system effectively reduces waste classification errors and significantly improves waste management efficiency.