Skip to content

Latest commit

 

History

History
48 lines (40 loc) · 1.7 KB

install.md

File metadata and controls

48 lines (40 loc) · 1.7 KB
Raw

Installation

Prerequisite:

If you want to use only semantic cloud segmentation node just build the package in a catkin workspace, for example:

mkdir -p ~/catkin_ws/src
cd ~/catkin_ws/src
git clone https://github.com/ctu-vras/traversability_estimation
git clone https://github.com/ctu-vras/cloud_proc
cd ~/catkin_ws/
rosdep install --from-paths /catkin_ws --ignore-src --rosdistro noetic -y
catkin build

In case you would like to run geometric cloud segmentation, traversability fusion or image segementation to point cloud projection nodes, please follow the extended proceedure (requires access to another repositories):

  • Install vcstool for workspace creation: 
    sudo apt install python3-vcstool
  • Create and build ROS workspace: 
    cd ~/catkin_ws/
vcs import src < src/traversability_estimation/config/workspace.repos
catkin config -DCMAKE_BUILD_TYPE=Release
catkin build

Put the weights to ./config/weights/ folder:

./config/weights/
  ├── hrnetv2_w48_imagenet_pretrained.pth
  ├── seg_hrnet_ocr_w48_train_512x1024_sgd_lr1e-2_wd5e-4_bs_12_epoch484/
  ├── depth_cloud/
  └── smp/
      └── se_resnext50_32x4d_352x640_lr1e-4.pth

One may also download datasets to train images and point cloud segmentation models. Please, refer to ./docs/rellis.md or ./docs/trav_data.md for examples.