An Automated System for Waste Segregation Using Deep Learning Algorithms

In current society, the issue of a proper waste disposal system is one of the major concerns for a clean and green environment. The most common method of waste disposal has always been dumping and burning regardless of the waste category. A better method of waste disposal can be implemented, if waste is categorically separated and is then disposed of using various methods, and if salvageable then may be reused or recycled thus reducing wastage and improving sustainability. Traditional automated waste detection systems only employ sensors that cannot accurately categorize waste into the ideal category for proper waste disposal. This paper proposes a waste detection system that uses image processing, categorizes waste into a few categories and then segregates them as per the division assigned to them. A large dataset is used to train a Convolutional Neural and then waste is identified by applying image processing techniques. The object then moves down the line on a conveyor belt after it is identified and a corresponding shaft is activated which pushes the object down to the bucket assigned to the waste category. The entire system is integrated using an Arduino and is remotely controlled. This system is used to facilitate simplifying the process of recycling resources that may have been disposed of and then turn back these resources into reusable resources. This will improve the process of waste segregation and disposal and automate end to end process as opposed to having any human involvement thus improving the efficiency. The CNN model is trained on a dataset with 5 main categories and gives an accuracy of 93%.