https://github.com/arcticray/crack-detection

A python machine learning project aimed at automatically identifying cracks on concrete surface images using a Convolutional Neural Network.
https://github.com/arcticray/crack-detection

cnn-classification python

Last synced: about 1 year ago
JSON representation

A python machine learning project aimed at automatically identifying cracks on concrete surface images using a Convolutional Neural Network.

• Host: GitHub
• URL: https://github.com/arcticray/crack-detection
• Owner: ArcticRay
• Created: 2024-11-28T21:35:45.000Z (over 1 year ago)
• Default Branch: main
• Last Pushed: 2025-02-03T16:38:45.000Z (over 1 year ago)
• Last Synced: 2025-05-15T19:13:35.034Z (about 1 year ago)
• Topics: cnn-classification, python
• Language: Python
• Homepage:
• Size: 547 KB
• Stars: 1
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

          
# Surface Crack Detection

![Project Banner](outputs/sample_images.png)

## 📄 Description

**Surface Crack Detection** is a machine learning project aimed at automatically identifying cracks on surface images.

**Confusion Matrix**

![Project Banner](outputs/confusion_matrix.png)

## 🚀 Features

- **Data Preprocessing:** Loading, resizing, and normalizing images.

- **Model Training:** Development and training of a CNN for image classification.

- **Model Evaluation:** Assessing model performance with metrics and visualizations.

- **Image Classification** Uploading and classifying

- **Explainability (LIME):** Generating visual explanations showing which regions of an image the model relies on when predicting cracks.

**Example LIME Explanation**  

Below is a sample visualization using LIME, highlighting important superpixels contributing to the “Positive” (crack) classification:

![LIME Explanation](outputs/lime/lime_explanation.png)

## Dataset

The dataset comprises images categorized into two classes:

- **Positive:** Images containing surface cracks.

- **Negative:** Images without surface cracks.

Özgenel, Çağlar Fırat (2019), “Concrete Crack Images for Classification”, Mendeley Data, V2,

[doi: http://dx.doi.org/10.17632/5y9wdsg2zt.2x](https://data.mendeley.com/datasets/5y9wdsg2zt/2)