Lesion Segmentation With Histogram Based Thresholding

This is the code of the project "Lesion Segmentation With Histogram Based Thresholding", created by: Francisco Nicolás Noya Cecilia Szambien

In the context of the course IMA201 at Télécom Paris 27/11/2024.

Project Demo

Project Description

This project focuses on image segmentation for skin lesions. The primary objective is to develop a method that, given an image of a skin lesion, generates a mask isolating the lesion from the surrounding skin.

Project Dependencies

The dependencies of this project can be found in the file requirements.txt.

To install the requirements you must use the following commando:

pip install -r requirements.txt

Project Structure

This project is structures in the following folders:

• manager: This folder contains the Manager class which manages the communication between classes and the adjust_hyperparamenters file which look for the hyperparameters of the model that optimise the metrics.
• pre_processing: This folder contains all the classes related to the preprocessing stage, sucha as: circular filter, hair removal, channel extraction, intensity adjustement and median filter.
• presentations: This folder contains the presentations given to the professor when where needed.
• research: This folder contains the research, which consists on both, theoretical and practical research.
• segmentation: This folder contains the segmentation class that is in charge of the segmentation and postprocessing steps of the model.
• dataset: Even thought this folder might not appear in the gitlab project, it contains the data used to test the model's performance and also adjust the hyperparameters.
• demo: The demo folder contains a Streamlit program that performs image segmentation on an input image.

Methodology of work

The methodology of work was as follows:

• There was a main branch that contains the last stable version.
• Every time someone wants to work, it has to create a branch from main.
• After working it should test that everything works and commit it work to the created branch.
• If there were no conflicts, then the created branch could be merge with main.
• If there were conflicts, then the created branch must be rebased from main and, after solving the conflicts, it will be merged to main.
• There was no compulsory peer review but it was encourage.