Detect Scratches

Experiments into the field of computer vision done at my three month internship at Rongheng Information Technology Co. Ltd. in Shanghai.

Requirements:

• numpy
• matplotlib
• Pillow
• scipy

see docker/requirements-full.txt for the full list

Python only deployment:

• Clone the repository: git clone https://github.com/felixschloesser/DetectScratches.git
• Enter the directory: cd DetectScratches/docker
• Download the dependencies using pip: pip install -r requirements-full.txt
• Make sure the images you want to inspect are in the test-images/directory
• Execute using ./src/main.py

You can use the -o flag to output the result images.

Docker deployment:

Clone the repository using git clone https://github.com/felixschloesser/DetectScratches.git

Enter the directory: cd DetectScratches/docker

All you have to do is to adjust the binding path inside the docker-compose.yml to your cloned notebook folder.

Docker will take care of the rest for you :) docker-compose up Now you can navigate to http://127.0.0.1/:8888 to access the jupyter notebook.

Have fun!

---

Documentation

Development Environment: Jupyter Notebook

Required Libaries:

• lib/image.py: doc
• lib/edge.py: doc
• lib/crop.py: doc

Unused Libraries:

• lib/signal.py

External Libraries:

For questions about external libraries, please read the corresponding documentation. I have linked them below.

• NumPy: docs
• scipy: docs
• Pillow: docs
• matplotlib: docs
• scikit-image: docs
• ipywidgets: docs

Literature

• Programming Computer Vision, O'Reilly, ISBN: 9781499316539, https://github.com/jesolem/PCV
• Hands-On Image Processing with Python, O'Reilly, ISBN: 9781789343731