Development Of Waste Segregator To Enhance Waste Classification Based On Deep Learning Approach

Rajandran, Venothraaj (2018) Development Of Waste Segregator To Enhance Waste Classification Based On Deep Learning Approach. Project Report. UTeM, Melaka, Malaysia. (Submitted)

Text (24 Pages) Development Of Waste Segregator To Enhance Waste Classification Based On Deep Learning Approach.pdf - Submitted Version Download (338kB)

Abstract

Malaysia is certainly one of the most successful nations in the changeover. The growth in industrialization and human population causes an increment in the solid waste materials. The intention of this research is to find an alternative way for waste disposal and develop a semi-automated recycling process. The common recycling method is manually picking and sorting the solid waste into categories, but due to the high cost to manage the waste generation and limited access to the recycling bin, deep learning based waste classification technique was proposed to solve this problem. The extent of this project is to use the depthwise separable convolutional neural network (MobileNets) for the plastic bottle and aluminium can classification. The system is also supported by Raspberry Pi and also python programming language. As a summary, the MobileNets based object detection is used to develop the waste segregator and able to achieve about 90% rate of success of waste detection. Aluminium can have a higher detection accuracy compare to the plastic bottle due to the transparency and also light reflectivity. The detection and segregation able to work as standalone but the aim to make fully automated is still not achieved.

Actions (login required)

View Item