This repository contains the source code for the paper Toward 3D Object Reconstruction from Stereo Images.
Important Note: The source code is in the (Stereo2Voxel/Stereo2Point) branches of the repository.
@article{xie2021towards,
title={Toward 3D Object Reconstruction from Stereo Images},
author={Xie, Haozhe and
Tong, Xiaojun and
Yao, Hongxun and
Zhou, Shangchen and
Zhang, Shengping and
Sun, Wenxiu},
journal={Neurocomputing},
year={2021}
}
We use the StereoShapeNet dataset in our experiments, which is available below:
The pretrained models on StereoShapeNet are available as follows:
- Stereo2Voxel for StereoShapeNet (309 MB)
- Stereo2Point for StereoShapeNet (356 MB)
git clone https://github.com/hzxie/Stereo-3D-Reconstruction.git
cd Stereo-3D-Reconstruction
pip install -r requirements.txt
git checkout Stereo2Voxel
git checkout Stereo2Point
cd extensions/chamfer_dist
python setup.py install --user
You need to update the file path of the datasets:
__C.DATASETS.SHAPENET.LEFT_RENDERING_PATH = '/path/to/ShapeNetStereoRendering/%s/%s/render_%02d_l.png'
__C.DATASETS.SHAPENET.RIGHT_RENDERING_PATH = '/path/to/ShapeNetStereoRendering/%s/%s/render_%02d_r.png'
__C.DATASETS.SHAPENET.LEFT_DISP_PATH = '/path/to/ShapeNetStereoRendering/%s/%s/disp_%02d_l.exr'
__C.DATASETS.SHAPENET.RIGHT_DISP_PATH = '/path/to/ShapeNetStereoRendering/%s/%s/disp_%02d_r.exr'
__C.DATASETS.SHAPENET.VOLUME_PATH = '/path/to/ShapeNetVox32/%s/%s.mat'
To train GRNet, you can simply use the following command:
python3 runner.py
To test GRNet, you can use the following command:
python3 runner.py --test --weights=/path/to/pretrained/model.pth
This project is open sourced under MIT license.
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent