Skip to content
"Conditional Single-view Shape Generation for Multi-view Stereo Reconstruction" in Tensorflow
Python C++ Cuda Shell
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Type Name Latest commit message Commit time
Failed to load latest commit information.
config update Apr 20, 2019
functions debug Apr 17, 2019
lib polish Apr 23, 2019 update link. Nov 29, 2019 polish Apr 23, 2019 add Apr 20, 2019 update. Apr 20, 2019
requirements.txt add requirements.txt Apr 12, 2019 update README and shell Apr 20, 2019


Created by Yi Wei, Shaohui Liu and Wang Zhao from Tsinghua University.


This repository contains source code for Conditional Single-view Shape Generation for Multi-view Stereo Reconstruction in tensorflow.

prediction example


The code has been tested with Python 2.7, tensorflow 1.3.0 on Ubuntu 16.04.

1. Clone code

git clone

2. Install packages

Python virtual environment is recommended.

cd OptimizeMVS
virtualenv env
source ./env/bin/activate
pip install -r requirements.txt

You need to compile the external libraries:



1. One-button setup


2. Two-stage training


Default training options are stored in the config folder.

3. Evaluation

You can download our pretrained model: single-category, multi-category and move it to a folder named demo for evaluation.


To evaluate your own model, set the 'load' option in as the path to your model.


The results might be slightly better than reported in the paper.


Part of the external operators are borrowed from latent_3d_points and PointNet++. The multi-view images were rendered from ShapeNetCore with the preprocessing scripts in mvcSnP and the point cloud data was from latent_3d_points. We sincerely thank the authors for their kind help.

This work was supported in part by the National Natural Science Foundation of China under Grant U1813218, Grant 61822603, Grant U1713214, Grant 61672306, and Grant 61572271.


If you find this work useful in your research, please consider citing:

  author = {Wei, Yi and Liu, Shaohui and Zhao, Wang and Lu, Jiwen and Zhou, Jie},
  title = {Conditional Single-view Shape Generation for Multi-view Stereo Reconstruction},
  booktitle = {Proceedings of IEEE Conference on Computer Vision and Pattern Recognition (CVPR)},
  year = {2019}

The first three authors share equal contributions.

You can’t perform that action at this time.