This library is a simple demonstration of using Optimal Transport Mapping Estimation [1] in a context of seamless copy between images. See [1] for details over the method
The Library has been tested on Linux and MacOSX. Among other classical dependencies, it requires the installation of POT, the Python Optimal Transport library (https://github.com/rflamary/POT)
- Numpy (>=1.11)
- Scipy (>=0.17)
- Matplotlib (>=1.5)
- Pyamg (>=3.1)
- POT (>=1.0)
If you want to execute the video demo, then you also need to have OpenCV for python installed.
One notebook is provided as example of use:
The video demo is available in the test_video.py file.
[1] M. Perrot, N. Courty, R. Flamary, A. Habrard, "Mapping estimation for discrete optimal transport", Neural Information Processing Systems (NIPS), 2016.