ROSMRF: Real-time Online 3D Semantic segmentation for mobile scene via Multi-modal Representation and Fusion
System : ubuntu 16.04
Hardware: Quadro RTX 6000
Software: Pytorch 1.3.0, CUDA-10.0, python 3.5+ For more packages information, we prepared them in the requires.txt file.
1.clone the SSF-DAN repository:
git clone https://github.com/Biotan/ROSMRF
2.prepare environment and compile the repository:
virtualenv -p /usr/bin/python3.5 ~/virtualenv/ROSMRF
source ~/virtualenv/ROSMRF/bin/activate
pip install -r requires.txt
bash develop.sh
3.download the row data of S3DIS dataset and use the script to preprocess:
cd S3DIS/
python preprocess.py
python prepare_data.py
4.train the network
python unet.py
5.evaluate the network
python unet_val.py
6.visualization the network
python unet_vis.py
| 
| 
|