Auto-WCEBleedGen Challenge

Automatic Detection and Classification of Bleeding and Non-Bleeding frames in Wireless Capsule Endoscopy

Note : can download trained models from latest releases

Table of Contents

• Introduction
• Evaluation Results
  • Classification Report
  • Detection
• Validation Dataset Results
  • Classification
  • Detection
  • Interpretability Plots
• Testing Dataset Results
  • Test Dataset 1
    • Classification
    • Detection
  • Test Dataset 2
    • Classification
    • Detection

Introduction

This README provides an overview of the Auto-WCEBleedGen Challenge project, which focuses on the automatic detection and classification of bleeding and non-bleeding frames in wireless capsule endoscopy.

Evaluation Results

Classification Report

These results are evaluated using validation data.

	Precision	Recall	F1-Score	Support
Bleeding	1.00	1.00	1.00	258
Non-Bleeding	1.00	1.00	1.00	265
Accuracy			1.00	523
Macro Avg	1.00	1.00	1.00	523
Weighted Avg	1.00	1.00	1.00	523

Detection

Metric	Value
Average Precision	0.81742
Mean Average Precision	0.57529
Intersection over Union	0.67174

Validation Dataset Results

Classification Screenshots

Detection Screenshots

Interpretability Plots

Testing Dataset Results

Test Dataset 1

Classification

Detection

Test Dataset 2

Classification

Detection