synesthesiax

ROS 2 package that projects LiDAR point clouds onto semantic camera images and publishes colored semantic point clouds.

Refactored node structure

The package now builds a single reusable executable:

synesthesiax_node

The same executable is launched twice with different parameters:

• synesthesiax_front
• synesthesiax_back

Use:

ros2 launch synesthesiax synesthesiax.launch.py

Single-side launch files are also available:

ros2 launch synesthesiax synesthesiax_front_camera.launch.py
ros2 launch synesthesiax synesthesiax_back_camera.launch.py

Important parameters

• labels_transport: raw or compressed.
• semantic_cloud_topic: output topic for the full semantic cloud.
• overlay_topic: output topic for the projected debug image.
• class_cloud_topic_prefix: prefix for per-class semantic clouds.
• enable_range_filter: enable/disable range filtering.
• enable_fov_filter: enable/disable FOV filtering.
• require_positive_x: discard points with x <= 0 when enabled.
• sync_queue_size: approximate-time synchronizer queue size.

Coordinate convention

The old back-camera implementation manually projected points as:

(-x, -y, z)

That convention is now absorbed in config/pinhole_model_back_camera.yaml:

R_new = R_old * diag(-1, -1, 1)

Therefore, the C++ projector is now frame-agnostic and always uses the original cloud coordinates directly.