CROWDBOT_perception

This is the perception pipeline for the CROWDBOT project, featuring person detection and tracking from multi-sensor modalities. The project is based on earlier works from the SPENCER project[repo], [repo]. New features include:

• Adaption for operating on Jetson AGX
• YOLOv3 with TensorRT acceleration for fast RGB-based person detection
• DROW3/DR-SPAAM for person detection with 2D LiDAR sensors
• [OPTIONAL] Optical flow aided tracking
• [OPTIONAL] 3D person pose analysis
• [OPTIONAL] Person ReID

For setting up, refer to rwth_crowdbot_launch/README.md.

Acknowledgement

• SPENCER tracking pipeline [repo], [repo]
• YOLOv3 with TensorRT
• darknet_ros for person detection
• DROW3 and DR-SPAAM for 2D-LiDAR-based person detection
• MeTRAbs for monocular 3D pose estimation
• pwc_net for optical flow estimation