krentzd / cpd-net-1

Coherent Point Drift Networks: Unsupervised Learning of Non-Rigid Point Set Registration (CPD-Net). Lingjing Wang, Xiang Li, Jianchun Chen, Yi Fang.

CPD-Net

(1) The code is tested on: python 3.6

It requires packages: pytorch and tensorflow. We use a wrapper sugartensor for constructing network. It can be easily downloaded at: pip install sugartensor.

(2) For training the model: python train.py

(3) For testing the model: python test.py

(4) C.D. (Before and After) will be printed out after running test.py and the qualitative result for visulization is saved at result_visulization folder.

(5) For comparison with CPD, we use the package: https://github.com/siavashk/pycpd

(6) geotnf is modified based on the code from paper: Convolutional neural network architecture for geometric matching,Ignacio Rocco, Relja Arandjelović, Josef Sivic.