Repository files navigation

Shubham Tulsiani, Tinghui Zhou, Alexei A. Efros, Jitendra Malik. In CVPR, 2017.Project Page

Please check out theinteractive notebook which shows reconstructions using the learned models. You'll need to -

• Install a working implementation of torch and itorch.
• Download the pre-trained models forPascal3D (490MB) andShapeNet (250MB). Extract the pretrained models to 'cachedir/snapshots/{pascal,shapenet}/'
• Edit the path to the blender executable in the demo script.

To use our proposed loss function for training, we need to compile the C implementation so it can be used in Torch.

cd drcLossluarocks make rpsem-alpha-1.rockspec

For training your own models and evaluating those, or for reproducing the main experiments in the paper, please see the detailed README files forPASCAL3D orShapeNet.

You'll need to install some additional dependencies (json and matio).

sudo apt-get install libmatio2luarocks install matioluarocks install json

If you use this code for your research, please consider citing:

@inProceedings{drcTulsiani17,  title={Multi-view Supervision for Single-view Reconstruction  via Differentiable Ray Consistency},  author = {Shubham Tulsiani  and Tinghui Zhou  and Alexei A. Efros  and Jitendra Malik},  booktitle={Computer Vision and Pattern Regognition (CVPR)},  year={2017}}