Polyp Segmentation in Endoscopic Images

Overview

This repository contains the code, models, and research documentation for automating polyp segmentation in gastrointestinal endoscopy images. Accurate and early detection of polyps is critical for the prevention of colorectal cancer. This project utilizes deep learning-based computer vision techniques to generate precise segmentation masks for medical image analysis.

Research & Documentation

Read the Full Paper: Click the thumbnail below to view the full research paper directly in GitHub's native PDF viewer.

Methodology & Architecture

• Task: Semantic Segmentation
• Architecture: U-Net, ResUNet, Attention U-net
• Loss Function: Dice Loss, IoU
• Data Processing: The image preprocessing has been mentioned in the paper under the preprocessing section.

📊 Key Results

The model was evaluated using standard segmentation metrics:

• Dice Coefficient: 0.9525%
• Intersection over Union (IoU): 0.9115

Visual Predictions

Original Image	Ground Truth Mask	Modified Mask	Model Prediction

⚙️ Installation & Usage

Since this is an archived research project, you can easily replicate the environment to run inference.

1. Clone the repository:

   git clone [https://github.com/yourusername/your-repo-name.git](https://github.com/yourusername/your-repo-name.git)
   cd your-repo-name