Installation

General Requirements

This codebase is tested with torch==1.10.0 and torchvision==0.11.0, with CUDA 11.3 and gcc 7.3.0. In order to successfully reproduce the results reported in our paper, we recommend you to follow the exact same configuation with us. However, similar versions that came out lately should be good as well.

Step 1: Create Enviroment

conda create -n pcseg python=3.7

Step 2: Activate Enviroment

conda activate pcseg

Step 3: Install PyTorch

conda install pytorch==1.10.0 torchvision==0.11.0 cudatoolkit=11.3 -c pytorch -c conda-forge

Step 4: Install Necessary Libraries

4.1 - nuScenes devkit

:oncoming_automobile: Note: This toolkit is required in order to run experiments on the nuScenes dataset.

pip install nuscenes-devkit 

4.2 - PyTorch Scatter

conda install pytorch-scatter -c pyg

4.3 - TorchSparse

Note: The following steps are required in order to use the voxel and fusion backbones in this codebase.

• Make a directory named torchsparse_dir

cd package/
mkdir torchsparse_dir/

• Unzip the .zip files in package/

unzip sparsehash.zip
unzip torchsparse.zip

mv sparsehash-master/ sparsehash/

• Setup sparsehash (Note that ${ROOT} should be your home path to the PCSeg folder)

cd sparsehash/
./configure --prefix=/${ROOT}/PCSeg/package/torchsparse_dir/sphash/

make

make install

• Compile torchsparse

cd ..
pip install ./torchsparse

• It takes a while to build wheels. After successfully building torchsparse, you should see the following:

Successfully built torchsparse
Installing collected packages: torchsparse
Successfully installed torchsparse-1.4.0

4.4 - Range Image Library

cd package/

• Unzip the range_lib.zip file in package/

unzip range_lib.zip
cd range_lib/
python setup.py install

• After successfully building range_lib, you should see the following:

Processing dependencies for rangelib==1.0.0
Finished processing dependencies for rangelib==1.0.0

4.5 - Other Packages

pip install pyyaml easydict numba torchpack strictyaml llvmlite easydict scikit-image tqdm SharedArray prettytable opencv-python

pip uninstall setuptools
pip install setuptools==59.5.0

4.6 - Register PCSeg

Inside PCSeg directory:

python setup.py develop