nalankaru/cs284-image-segmentation

Image Segmentation using Conditional Random Fields

cs284-image-segmentation

A python application which performs image segmentation of the Weizmann Horse Dataset using Conditional Random Fields (CRF). A Mini Project requirement for CS 284 course at University of the Philippines Diliman AY 2016-2017 under Sir Prospero Naval.

Requirements

• pystruct
• OpenCV Computer Vision Library
• Scikit Image
• Scikit Learn
• Python 2.7.12

Instructions

1. Clone and download the repository.
2. Train the CRF model.

python -W ignore imagesegmentation.py

3. Update the horseModelLogger variable with the generated model file. Run the demo application to visualize segmentation.

python -W ignore imagesegmentationDemo.py WeizmannSingleScale/horses/training/images/image-xxx.png

python -W ignore imagesegmentationDemo.py WeizmannSingleScale/horses/test/images/image-xxx.png

4. Measure the pixelwise performance of the model.

 python -W ignore imagesegmentationPerf.py