|
Proper waste segregation is a crucial task for waste management at the household level, yet it is often an error-prone process. According to Repak, in Ireland 100,000 tonnes of recyclable material is sent to landfill every year due to contamination. This project aimed to create a simple device that uses convolutional neural networks (CNN) and robotics to eliminate the problem from the hands of the user, for reduced contamination rates.
The CNN was trained on 1000+ images sourced from Kaggle and specific object images were taken locally. Using transfer learning, it achieved an accuracy of 96%, showing that the robot can accurately classify waste materials and perform sorting tasks with high efficiency. The robot utilizes an ultrasonic sensor to detect the presence of objects in front of it and a Raspberry Pi running the CNN to classify waste materials into recyclable, general waste, or organic waste categories. The robot's movement is controlled using small servos, and the data collected from the sensor is processed using the Raspberry Pi's GPIO pins. The robot is designed to be portable and wall mountable to accommodate different bin sizes.
The proposed system aims to reduce human involvement in waste sorting and promote efficient waste management practices. Overall, this project provides a promising solution for automating waste sorting and reducing the rate of recycling contamination at the household level.
|