LIX-shape-analysis / SURFMNet

code and data associated with the paper "Unsupervised Deep Learning for Structured Shape Matching"
50 stars 7 forks source link

SURFMNet

Source code and data associated with the ICCV'19 oral paper "Unsupervised Deep Learning for Structured Shape Matching". A cleaner and object oriented code is available at https://github.com/Not-IITian/SURFMNet-Object_oriented

Dependency

The code is tested under TF1.6 GPU version and Python 3.6 on Ubuntu 16.04, with CUDA 9.0 and cuDNN 7. It requires Python libraries numpy, scipy.

Download Pre-processed Mesh Data

Please run bash Prepare_data.sh

Shape Matching

To train a DFMnet model to obtain matches between shapes without any ground-truth or geodesic distance matrix (using only a shape's Laplacian eigenvalues and eigenvectors and also Descriptors on shapes):

    python train_DFMnet.py

To obtain matches after training for a given set of shapes:

    python test_DFMnet.py

Visualization of functional maps at each training step is possible with tensorboard:

    tensorboard --logdir=./Training/

Download GT Correspondence and precomputed pairwise matches for some baselines

https://drive.google.com/open?id=1qvqtJz-_zvMxC0ZMuFGbtlKxc9Py3Ggg

Download Geodesic Matrices for Faust and Scape remesh from here:

https://www.dropbox.com/s/ryvc1b0c3gnz2ju/Faust_r_test.zip?dl=0 https://www.dropbox.com/s/ysrctegmqgpo72z/scape_test.zip?dl=0

For any further question, please send an email to Abhishek at kein.iitian@gmail.com.