This project aims to automate the segmentation of tumors from lung scans using the UNet Architecture.
- The preprocessing folder includes exploratory data analysis, general preprocessing and some experimental advanced preprocessing techniques
- The UNet folder contains model training and a notebook which explains how to use it for inference
Model can be found at: https://www.kaggle.com/models/aryankashyapnaveen/u-net-lung-segmentation
We have used the following paper as reference: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10451633/
- Use the mobile-net backbone.
- Improve performance with advanced preprocessing techniques.