AutoWCEBleeding Challenge

Introduction

Welcome to my submission for the Auto-WCEBleedGen challenge. The goal is to develop and evaluate an AI model for the automatic detection and classification of bleeding and non-bleeding frames in Wireless Capsule Endoscopy (WCE) Images.

Repository Contents

Flow

The directory contains two sub-directory for the classification and segmentation.
- First the image is classified into bleeding and non-bleeding image.
- Then, the trained model is used to detect the bleeding region within the bleeding image.

Code

The code is divided into two submodules: classification and segmentation.

Classification

• train.py: Python script for training our AI model.
• validate.py: Python script for validating our AI model.
• test.py: Python script for testing our AI model.
• utils/: Directory containing utility scripts and helper functions.
• config/: Directory containing configuration files.
• checkpoints/: Directory to store model checkpoints (optional).
• assets/: Directory for additional assets and resources.
• README.md: This README file.

Segmentation

Model

• [Include details about your trained model here, if applicable.]

Excel Sheet

• predictions.xlsx: Excel sheet containing image IDs and predicted class labels for testing dataset 1 and 2.

Usage

• Classification
• Configure the config file.
• Prepare the data by running the data_preparation.py file.
• Use train.py to train the model on the data.
• Next, evaluate the performance using evaluate.py file.

Evaluation Metrics

Here are the evaluation metrics for the model:

• Classification

• Detection

Future Work / To-Do List

• ~~Segregate data~~
• ~~Add training scripts.~~
• ~~Add configuration file.~~
• Add evaluation scripts.
• Add visualization script
• Develop pipeline for inference.
• Update the readme.
• Add images and final compiled result.

Questions or Feedback

If you have any questions or feedback, please feel free to contact us.