Dense Human Correspondence via Learning Transformation Synchronization on Graphs

Pytorch implementation of Our Neurips 2020 paper

Dense Human Correspondence via Learning Transformation Synchronization on Graphs (pdf)

We demonstrate the steps to reproduce our results as well as providing pre-computed correspondence results and error statistics.

Dependencies

Reproducing Experimental Results on Human Shapes

Here we demonstrate the full pipeline to reproduce our approach.

Pre-computed Correspondence Results and Error Statistics

Here we provide a convenient link to download pre-computed results on SHREC19-Human and FAUST dataset.

Reproducing Experimental Results on Animal Shapes

This is an additional experiment to test our approach's generalizability on non-human shapes.

Contact

Please send any questions regarding this code to Xiangru Huang (xiangruhuang816 [at] gmail [dot] com).