Skip to content

Latest commit

 

History

History
15 lines (10 loc) · 949 Bytes

README.md

File metadata and controls

15 lines (10 loc) · 949 Bytes
Raw

Work in progress

  1. Start camera, CSI: rosrun jetbot_ros jetbot_camera or RealSense: roslaunch realsense2_camera rs_camera.launch

  2. Calculate camera homography using the estimator:

  • Print the pattern
  • Place the robot in front of the pattern
  • Run the calibration: roslaunch drive_ros_camera_homography homography_estimator.launch and use dynamic_reconfigure to tune the parameters.
  • Frames for the pattern should be bigger than the pattern (didn't work for me otherwise).
  • Homography will be saved into the current folder, see example in the repo.

  1. Use DCNN inference for semantic segmentation to get the mask, like rosrun semantic_segmentation segmentation_node __ns:=/jetbot_camera _topic_image:=camera

  2. TODO: Apply homography to the mask from segmentation and convert it into the local costmap