This is the official code repository for FlashFusion, an efficient dense 3D reconstruction system that only relies on CPU computing.
This is a project from LuVision SIGMA, Tsinghua University. Visit our website for more interesting works: http://www.luvision.net/
This project is released under the GPLv3 license. We only allow free use for academic use. For commercial use, please contact us to negotiate a different license by: fanglu at tsinghua.edu.cn
If you find our code useful, please kindly cite the following papers:
@inproceedings{han2018flashfusion,
title={FlashFusion: Real-time Globally Consistent Dense 3D Reconstruction using CPU Computing.},
author={Han, Lei and Fang, Lu},
booktitle={Robotics: Science and Systems},
volume={1},
number={6},
pages={7},
year={2018}
}
@ARTICLE{8606275,
author={Han, Lei and Xu, Lan and Bobkov, Dmytro and Steinbach, Eckehard and Fang, Lu},
journal={IEEE Transactions on Robotics},
title={Real-Time Global Registration for Globally Consistent RGB-D SLAM},
year={2019},
volume={35},
number={2},
pages={498-508},
doi={10.1109/TRO.2018.2882730}}
- Ubuntu 16.04/18.04
- Intel cpu
source prepare.sh
mkdir build
cd build
cmake ..
make -j
The format of offline dataset follows TUM RGBD Dataset.
Example sequences can be downloaded at http://221.228.239.253:81/opensource_data/FlashFusion/dataset.zip
Put the dataset at dataset/
, organized as
dataset
|---- iclnuim
|---- tum
|---- xtion
sudo ./FlashFusion sensor_type calib_file voxel_size dataset_path
sensor_type: 0:dataset 1:realtime openni_camera(such as xtion) 2: realtime realsense_camera
Example:
sudo ./FlashFusion 0 ../param/calib_tum.txt 0.005 ../dataset/tum/fr1_desk/
sudo ./FlashFusion 0 ../param/calib_icl.txt 0.005 ../dataset/iclnuim/icl0n/
sudo ./FlashFusion 0 ../param/calib_xtion.txt 0.005 ../dataset/xtion/
sudo ./FlashFusion 1 ../param/calib_xtion.txt 0.005