SCFusion
Authors: Shun-Cheng Wu, Keisuke Tateno, Nassir Navab and Federico Tombari
This is the back-end network of SCFusion repo [paper].
We follow the data structure as the edge-connect repository
Install
# cuda version: nvcc --version
conda create --name scfusion pytorch torchvision tensorboard pillow==6.1 cudatoolkit=YOUR_CUDA_TOOLKIT_VERSION pyyaml -c pytorch -c conda-forge
conda activate scfusion
Build Meshing tool
The meshing tool allows converting the input/output files into meshes
cd extension
python build.py install
- Train
python train.py --config /pth/to/config.yml
- test
python test.py --config /pth/to/config.yml
- sample
need to install meshing tool extension
python sample.py --config /pth/to/config.yml
- trace
export the trained model to be used on c++
python trace.py --config /pth/to/config.yml
The pre-trained model can be downloaded here.