Kvasir-SEG polyp segmentation using U-Net based models

This project focuses on semantic segmentation for the Vision and Cognitive Systems course at UNIPD.

Dataset

The Kvasir-SEG dataset used in this project can be accessed here: https://datasets.simula.no/kvasir/.

Models

This repository contains the implementation of the following models:

• Attention UNet
• UNet++

Additionally, the final project includes benchmark models such as U-Net and ResUNet, implemented by a colleague (code not included in this repository).

Results

Our experiments demonstrate that UNet++ with augmented data and deep supervision outperforms all other models.

Additional Resources

This repository also includes:

• The final project report.
• Presentation slides.

We achieved a grade of 29/30 for this project!