MinkowskiNet

Authors: [Shi-Sheng Huang], please contact with me via shishenghuang.net@gmail.com

This is a C++ Inteface of MinkowskiNet, tested in Ubuntu 16.04,The details can be found in the following CVPR 2019 paper

Christopher B. Choy, JunYoung Gwak, Silvio Savarese: 4D Spatio-Temporal ConvNets: Minkowski Convolutional Neural Networks. CVPR 2019: 3075-3084

2. Prerequisites

We have tested the library in 16.04, but it should be easy to compile in other platforms. A powerful computer (e.g. i7) will ensure real-time performance and provide more stable and accurate results.

C++11 or C++0x Compiler

We use the new thread and chrono functionalities of C++11.

CUDA, ONLY CUDA 10.1 is ok in Ubuntu 16.04

REQUIRED for all GPU accelerated code at least with cmake it is still possible to compile the CPU part without available at https://developer.nvidia.com/cuda-downloads

If you have installed CUDA 10.1, just change your cuda to CUDA 10.1 if more than one CUDA version is installed

Glog

sudo apt-get install libgflags-dev libgoogle-glog-dev

Python 3.6, ONLY python 3.6 is OK

sudo add-apt-repository ppa:deadsnakes/ppa
sudo apt update
sudo apt install python3.6
sudo apt-get install python3.6-dev

3. Compile

install python virtual environment

./python_env.sh

then you have install python3.6 virtual environment.

download the Res16UNet34C model weight for ScanNet dataset

wget https://node1.chrischoy.org/data/publications/minknet/MinkUNet34C-train-conv1-5.pth

Set the MINKOWSKINET_WEIGHT_FILE in the CMakeLists file

just set MinkUNet34C-train-conv1-5.pth path located in your computer.

Complile

mkdir build
cd build
cmake ..
make -j8

4. Usage

how to use

1. Run the test

cd bin
./test

More Info

If any problem, please check Spatio-Temporal Segmentation