Meeting notes

Feb 15

• usage of 2, grid maps. one for the occupancy grid and the other overlaying the occupancy grid map which has the semantic label. Using YOLO for classification and projecting it to ground level where 2D lidar can be used for detection.
• Make quadric slam or later versions runnable by Friday.
• Which semantic map is to be formed? 2D or 3D?
• How will such a semantic map help in localization? is it useful for any other tasks?
• How much computation power does each mapping method requires?
• Storage space required for each mapping method.
• The final implementation should be runnable on HSR.
• Access to work lab PC for RnD works.

Mar 15

• Focus on QuadricSlam. Understand how a quadrics is defined, equations for it and how to plot a quadrics.
• Debug the code, understand the workflow of the code and prepare a block diagram.
• Try to fix the data assosciator part to make QuadricSLAM work on TUM dataset.
• Try different detectors instead of default detectron-based detection.