This project contains my work from my one month internship as a Data Scientist in Koncern IT at Københavns Kommune. During my internship I had the opportunity to participate in a Waste detection project, where I was exposed in the latest techniques in computer vision with python. I had the chance to learn and use libraries such as OpenCv and Pillow, how to do Data Augmentation and create Synthetic data.
In this project I have enlarged a dataset of images using different Augmentation techniques such as
- transparent background
- image sharpening
- rotation
- perspective transformation
- wave transformation
- add different background
I used the YOLOv7, which was trained in Google colab, on data that I created using the above data Augmentation techniques and using only 5 categories (plastic, paper, metal, glass, cardboard).
Here are some results after training the model:
