charlesbrownk / plastic-detection

This is one of the high school artificial intelligence projects, where yolov5 and my custom dataset were used to train a model.

Plastic-detection

This is one of the high school artificial intelligence projects, where yolov5 and my custom dataset were used to train a model.

Motivation

Korea boasts a recycling rate well above the OECD average recycling rate. However, if you look at the real recycling rate on the other side, it is similar to the OECD average recycling rate. Experts say that the reason for this is the difficulty of classification due to the various types of plastics. Therefore, it is more economically beneficial to discard plastics than to recycle them. After understanding these problems, I thought, 'What if artificial intelligence classifies plastics?'

Perform

First, a total of four classes were set for PE, PET, PP, and PS. After that, pictures were taken and data were collected. Insufficient data were supplemented in AIHub to compose the data. And I finally composed my custom dataset by working on data labeling in Roboflow. Since the basic data is 2479 sheets, which is very small to proceed with model learning, it has been augmented to a total of 7377 sheets through the Data Augmentation process.

Dataset

If you want to download my yolov5 custom dataset, go to releases and download my dataset.

Training

Check my wiki how to train custom model!

License

© 2022, Plastic-detection. Released under GNU General Public License v3.0.

Plastic-detection repository is authored and maintained by @CharlesbrownK.