Custom implementation of “GrabCut”: interactive foreground extraction using iterated graph cuts

My implementation of Grabcut: Interactive foreground extraction using iterated graph cuts, ACM SIGGRAPH 2004, without border matting.

Dependencies

• Python 3.6
• OpenCV
• NumPy
• scikit-learn
• python-igraph

File desctiptions

• GMM.py – Gaussian mixture model
• grabcut.py – Core implementation of the algorithm.
• grabcut_opencv.py – A copy of the official OpenCV sample.
• messi5.jpg – A copy of OpenCV’s sample image.

Usage

python grabcut.py <filename>

If filename is not specified, the program will pick messi5.jpg to segment. For more details, please refer to instructions printed to the console after run the program.

Download

Source code on GitHub: GrabCut

Slide: grabcut