Skip to content

Latest commit

 

History

History
56 lines (44 loc) · 3.09 KB

tips_usage.md

File metadata and controls

56 lines (44 loc) · 3.09 KB

Tips for usage

General

  • Sequential/Parallel mode: gflag parallel_run
    • Running in parallel (default): parallel_run=1.
    • Running in sequential mode: parallel_run=0. Or, if using the example script, use the -s flag at commandline.
  • Log output in csv files: gflag log_output=true. Or, if using the example script, use the -log commandline argument. By default, log files will be saved in output_logs directory.

Loop Closure Detector

  • Enable Loop Closures: use_lcd=true gflag (disabled by default). Or, pass -lcd to the example script to turn on the LoopClosureDetector module:

./scripts/stereoVIOEUROC.bash -p "PATH_TO_DATASET/V1_01_easy" -lcd

Structural Regularities

  • You can run Kimera-VIO using structural regularities, as in our ICRA 2019 paper, by specifying the option -r:

./stereoVIOEuroc.bash -p "PATH_TO_DATASET/V1_01_easy" -r

3D Visualizer

  • The 3D Visualization window implements the following keyboard shortcuts (you need to have the window in focus: click on it):

    • Press t: toggle freezing visualization.
    • Press v: prints pose of the current viewpoint of the 3D visualization window.
    • Press w: prints to the terminal the size of the 3D visualization window.

    These last two shortcuts are useful if you want to programmatically set the initial viewpoint and size of the screen when launching the 3D visualization window (this is done at the constructor of the 3DVisualizer class).

    • Do not press q unless you want to terminate the pipeline in an abrupt way (see #74).
    • Press s: to get a screenshot of the 3D visualization window.
    • Press 0, 1, or 2: to toggle the 3D mesh representation (only visible if the gflag visualize_mesh is set to true).
    • Press a: to toggle ambient light for the 3D mesh (gflag visualize_mesh has to be set to true).
    • Press l: to toggle lighting for the 3D mesh (gflag visualize_mesh has to be set to true).
  • The 3D Visualization allows to load an initial .ply file. This is useful for example for the Euroc dataset V1_01_* where we are given a point-cloud of the scene. Note that the .ply file must have the following properties in the header, and its corresponding entries:

    • A vertex element: element vertex 3199068 (actual number depends on how many vertices has the ply file).
    • A face element: element face 0(actual number depends on how many faces has the ply file)
    • Properties may vary, as for now, any given .ply file will be displayed as a point-cloud, not a mesh (due to some weirdness in opencv).
    • For example, the ground-truth point-cloud provided in V1_01_easy dataset must be modified to have the following header:
      ply
      format ascii 1.0
      element vertex 3199068
      property float x
      property float y
      property float z
      property float intensity
      property uchar diffuse_red
      property uchar diffuse_green
      property uchar diffuse_blue
      element face 0
      property list uint8 int32 vertex_indices
      end_header