An Object-level SLAM Method Based on Ellipsoidal Representation in Dynamic Scenes.
video address at: https://www.bilibili.com/video/BV1jt421G7G9/?vd_source=737506656063f049ffdc96e35434c482/
Our environments: ubuntu18.04/GCC7.5/ GPU: RTX2080ti.
Our sourse code: coming soon.
System critical step:
Our environments: ubuntu18.04/GCC7.5/GPU: RTX2080ti.
Our code is tested on:
- CMake 3.10.0
- Eigen 3.3.7
- NVIDIA CUDA 10.2
- OpenCV 4.5.5
- pytorch 1.6.0
- pcl 1.8.1
- ceres 1.14
- Sophus 1.0
sudo apt-get install libvtk7-dev libsuitesparse-dev liblapack-dev libblas-dev libgtk2.0-dev
gflags 2.2.2 set 'BUILD_SHARED_LIBS' to ON
glog 0.6.0
boost 1.65
####### Build ceres
ceres 1.14.0
####### Build Opt
git clone https://github.com/niessner/Opt.git
as in dynamic fusion
Optional:
- Pangolin
mkdir build && cd build
cmake -DVISUALIZATION=ON ..
make -j8
We use -DVISUALIZATION=OFF/ON to switch visualization plug.