Perception
Overview of image-facing perception modules for detection, tracking, and structured target output.
Layer Role
The perception layer turns raw sensor data into structured observations that upper layers can consume directly.
In the current repository, it mainly covers:
- object detection
- target tracking
- auto-aim visual front-end logic
- general YOLO-style 2D detections
Current Directory Mapping
driver/ros2_hik_cameraperception/rm_auto_aimperception/yolo_detector
Current Modules
Interface Guidance
New perception modules should preferably follow these rules:
- reuse standard image-facing inputs such as
/image_rawand/camera_info - keep detection, tracking, and control outputs separated
- do not mix control semantics into pure detector messages
- generic detectors should stay decoupled from competition-specific task logic