jimeiyang / deepRotator

This is the code for NIPS15 paper "Weakly-supervised Disentangling with Recurrent Transformations for 3D View Synthesis"

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

deepRotator

This is the code for NIPS15 paper Weakly-supervised disentangling with recurrent transformations for 3D view synthesis by Jimei Yang, Scott Reed, Ming-Hsuan Yang and Honglak Lee.

Please follow the instructions below to run the code.

  1. Download the preprocessed chair data from "https://www.dropbox.com/s/q2bih317hyhbpm5/chairs_data_64x64x3_crop.mat?dl=0", and save it to the "./chair/data/" folder.
  2. Compile the caffe and matcaffe in "./caffe-cedn/" (Makefile.config needs to be adjusted according to your machine).
  3. Train the model by running the matlab scripts "train_chair_rotator_base.m" and then "train_chair_rotator_curriculum.m".
  4. The pre-trained RNN16 model for chair rotation can be downloaded from "https://www.dropbox.com/s/h7iiei53u2g1vvn/rnn_t16_model.tar.gz?dl=0".

The single-view chair rotation demo can be found in the Youtube:

IMAGE ALT TEXT HERE

Please contact "jimyang@adobe.com" if any questions.

About

This is the code for NIPS15 paper "Weakly-supervised Disentangling with Recurrent Transformations for 3D View Synthesis"


Languages

Language:C++ 53.9%Language:MATLAB 16.8%Language:HTML 15.9%Language:Python 7.1%Language:Cuda 3.7%Language:C 0.8%Language:CMake 0.7%Language:Makefile 0.5%Language:Shell 0.3%Language:CSS 0.2%Language:Mercury 0.0%