Image segmentation

Lab 3 project on DTU by Benjamin Banks(s234802), Bjørn Hagbarth(s234870) and Tobias Rodrigues Bjerre(s234823)

Gettings started

You need to have Python3 installed.

When installed, run the following to install the required python packages:

pip install -r requirements.txt

Run the acutal script by running:

python main.py

Unsupervised learning protocol

1. Do K-nearest once on the image
2. Use the edge_score to calculate the inital backgroundcentroid
3. Find the oposite, and make it a foregroudcentroid
4. Run K-nearest n number of times.

Code documentation

Dev tools

• The Python Image Preview VSCode extention can be used.
• The tests have been written using python unittest. It is helpfull to run the tests in debug mode.

Folder stucture

• The modules are code written to do the unsupervised learning.
• The services are meant support thing, handeling stuff like loading and converting.

Other

• If no dimension is writen, 1d is implied.