Skip to content

Plot statistics

Jump to bottom
matlabbe edited this page Nov 24, 2014 · 6 revisions

Introduction

This tutorial will show the power of GUI for online monitoring of the core process. The idea is that all statistics from the RTAB-Map algorithm are sent after every iteration to the GUI. These statistics can be plotted/viewed in different ways.

Details

  • Open RTAB-Map application.

  • Show the Statistics panel : Menu Window -> "Show view" -> Statistics.

  • Show the Posterior panel : Menu Window -> "Show view" -> A posteriori PDF.

  • Apply settings to the detector : Menu Edit -> "Apply settings to the detector".

  • You should see some tabs appearing at the bottom of the Statistics panel. Expand the Timing tab. You should now see a main window like this (vertical layout is set in the GUI settings of the Preferences dialog).

  • Create a new plot figure with "Total time" statistic in the "Timing" tab.

  • You can customize some attributes of the plot with a right-click on the figure. Here we set "Maximum items shown" to 50, "Show grid" and modified the Title, xLabel and yLabel.

  • Start the detection for some images. Data should gradually be added to the plot. After a while, you can have something like that. Pause the detection.

  • Now add a new statistic to the figure 1. Add "Memory update" and "Likelihood computation" of the "Timing" tab to figure 1.

  • Resume the detection for some iterations. You should see something like this (we disabled the grid).

  • For a better viewing of some curves, you may want to hide other curves. Click on the legend "Timing/Likelihood computation/ms".

  • Another thing you may want to do is to set a threshold in the figure. Lets create a new figure with the statistic "Highest hypothesis value". So we now have a figure 2.

 

  • Open the context menu of the figure and select "Add line" -> "Horizontal line ...".

  • A dialog appears and set a name for the threshold.

  • Another dialog appears to set the value.

  • The resulting figure 2.

  • One last thing is about the "A posteriori PDF" panel opened at the beginning. If the detection has run for enough time, you should see a curve with some dots. A dot represents a location in the Working Memory. The dot "-1" represents the "No loop closure" hypothesis. You can select a dot (only in graphics view mode... see right-click menu) to show up the related image corresponding to this loop closure hypothesis.

Clone this wiki locally