Real time waste material detection using region-based convolutional neural network (RCNN) / Mohammad Aiman Haziq Mohd Hudzir

Mohd Hudzir, Mohammad Aiman Haziq (2019) Real time waste material detection using region-based convolutional neural network (RCNN) / Mohammad Aiman Haziq Mohd Hudzir. Degree thesis, Universiti Teknologi MARA (UiTM).
Abstract

Waste material could be classified into few categories whether it is plastic, paper, glass and others including general waste. Nowadays, there is an approach by providing recycle bin that sort the recyclable waste material into its own category. However, there could sometimes mistakes in sorting the waste material into a correct category. In this research, a real time waste material detection using Region-based Convolutional Neural Network (RCNN) is developed. Methodology in this system consist of dataset development, pre-processing, classification and evaluation. In dataset development, all the images of paper, plastic and glass will be collected and kept in a folder. Next, the dataset will undergo pre-processing that creates bound box around the images. Then it will be trained by using RCNN model. Tensorflow and Anaconda virtual environment will be setup. A camera will detect the object and display each of the object detected in a windows. Each object detected will have a bound box around it and a text showing what is the category. For the evaluation, the overall result from conduction several test is the prototype is capable of classification up to 99% accuracy at most and not lower than 66%. The result of the classification also can be increased by training more data set. This system will be a stand-alone type system so that it could be adaptive to any platform. Last but not least, this research can be expanded more in future by continuing the system implementation. The system is expected to be implemented into every recycle bin around the world.

Item Details
Edit Item
Edit Item
Downloads & Files
[thumbnail of 109970.pdf]
Text
109970.pdf
Download (80kB)
Location & Physical Holdings
Digital Copy
Physical Location
Item Status
Indexing & Metrics