K-Image segmenation using network flow

Both the C++11 and python dependencies are freely available. Installation is quite simple.

Code Dependency: C++ 11 C++ opencv 2.4.13

Python dependency: python version 2.7 scikit-learn numpy opencv

Compilation instructions:

If you have pkg-config for opencv, the following command should compile the file

2 Segmentation:

#Compilation: g++ -std=c++11 imagesegmentation.cpp -o imagesegmentation pkg-config --cflags --libs opencv

PS: If you don't have pkg-config you can expand the command pkg-config --cflags --libs opencv In MAC OSX the command expands as following. Might have to change your path depending on your installation!

-I/usr/local/Cellar/opencv/2.4.13/include/opencv -I/usr/local/Cellar/opencv/2.4.13/include -L/usr/local/Cellar/opencv/2.4.13/lib -lopencv_calib3d -lopencv_contrib -lopencv_core -lopencv_features2d -lopencv_flann -lopencv_gpu -lopencv_highgui -lopencv_imgproc -lopencv_legacy -lopencv_ml -lopencv_nonfree -lopencv_objdetect -lopencv_ocl -lopencv_photo -lopencv_stitching -lopencv_superres -lopencv_ts -lopencv_video -lopencv_videostab

To Run 2 segmenation, you can use 500500 which works within 34 minutes. Images under that size are pretty fast. The k-segmentation repeatedly runs network flow and hence is slower. Please run k-segmentation 200*200 (might take 7-8 mins )!

The downsample tool is avaialble to downsample the images python downsample.py

Just provide the imagename. No path. All images should be in the image folder integer value width of the image integer value of height of the image.

RUN:

Please provide the directly. Please don't provide the path. The images should be added into the image folder. The program automatically looks for images in the image folder.

./imagesegmentation

- name of the imagefile - an integer value 0 or 1, if you want to use gmm or k-means. It uses gmm by default.

K Segmenation :

g++ -std=c++11 kimagesegmentation.cpp -o kimagesegmentation pkg-config --cflags --libs opencv

Please look into the pkg-config expansion above.

RUN K-Segmentation

./kimagesegmentation

- name of the imagefile - an integer value 0 or 1, if you want to use gmm or k-means - The total number of segments you want the image to be divided into.