Given an image of a jumbled waste, we seek to categorize the different pieces of the waste into three categories: landfill, recycling, paper. This project utilizes Faster R-CNN to get region proposals and classify objects [6]. In this report, we will give an overview of our project and what we have done so far in terms of solving this problem. First, we define our waste sorting problem and present current research and solutions to similar object detection problems. We then give an outline of our proposed architecture and model for approaching our specific task, which entails using a fine-tuned Faster R-CNN. We will also describe the nature and generation of our dataset, as well as the results we achieved from our experiments. Lastly, we outline our next steps in terms of optimizing and improving upon our solution.
[1]
Gary Thung,et al.
Classification of Trash for Recyclability Status
,
2016
.
[2]
Luc Van Gool,et al.
The Pascal Visual Object Classes (VOC) Challenge
,
2010,
International Journal of Computer Vision.
[3]
Rob Fergus,et al.
Visualizing and Understanding Convolutional Networks
,
2013,
ECCV.
[4]
Andrew Zisserman,et al.
Very Deep Convolutional Networks for Large-Scale Image Recognition
,
2014,
ICLR.
[5]
Kaiming He,et al.
Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks
,
2015,
IEEE Transactions on Pattern Analysis and Machine Intelligence.