nfrumkin / SegmentationAlgorithms

A comparative view of traditional machine learning approaches with max-flow algorithms for simple foreground/background image segementation. A final project for EC504, Advance Data Structures and Algorithms

Segmentation Algorithms

Installation Instructions

Python Version = 3.6.6

Load Modules and Pip install

On SCC:

module load python3
pip3 install --user PyMaxFlow

note: Must use ssh -X for GUI!!

Python API Packages

• python-tk
• PIL
• numpy
• pickle
• scipy
• maxflow
• matplotlib
• skimage

Running Application

Step 1: Image Annotation

python3 image_gui.py

1. Click on foreground
2. Click-and-drag to annotate foreground of image
3. Click on background
4. Click-and-drag to annotate background of image
5. Click "Finish Labelling"
6. Close pop-up

Step 2: Max Flow Image Segmentor

python3 max-flow.py

Step 3: Otsu Thresholding

python3 otsu_thresholding.py