Unsupervised crack detection and more for microscpopic metal surfaces. Uses no deep/machine learning but just image processing from opencv.

Sample Output

Sample Report


Cracke has the following components:

Image Processing Pipeline (

  1. Median Blur Filtering
  2. Canny Edge Detection
  3. Morphological Segmentation
  4. Bounding Box Generation

Crack Estimation Pipeline (

  1. Approximate Crack Length 2 Use Paris Equation for crack propagation
  2. Compare with standard loading for a given material

Report Generation Pipeline (

  1. Status of the material
  2. Analysis of the image
  3. Estimate crack start and cause
  4. Generate report

Flask App (

Contains the flask app code for the HackaTUM 2019 demo presentation.

Cross Platform App (/my_app/)

Coontains code for the ionic app.


python dependencies - see 'requirements.txt' other dependencies: ionic.

Running the App

  • go to /app ionic cordova run browser

for server go to /server python

HackaTUM Presentation: Click Here.

Devpost Link: Click Here

PS: The entire pipeline was coded in a day for the hackathon and therefore is crude and raw and at times even unintelligeble. We'll try to fix bugs and clean up the code as much as possible. In the meanwhile please also feel free to fork and contribute to our codebase.


HackaTUM 2019

Built With

Share this project: