Vision#

rc.vision

BWSI Autonomous RACECAR Course Racecar Neo LTS

File Name: vision.py File Description: Defines the interface of the Vision module of the racecar_core library. Exposes Coral EdgeTPU object detections to student code as a list of Detection objects.

class vision.Detection(class_id: str, score: float, bbox: tuple)#

A single object detection result from the EdgeTPU.

class_id#: Label string (or numeric ID as string) of the detected object, as emitted by edgetpu_node.

score#: Confidence in [0.0, 1.0].

bbox#: (center_x, center_y, width, height) in pixels of the source image (the /camera/forward frame).

class vision.Vision#

Provides access to onboard ML object detection via the Coral EdgeTPU.

abstractmethod get_detections() → List[Detection]#

Returns the latest list of object detections from the EdgeTPU.

Returns:: A list of Detection objects with class_id, score, and pixel-space bounding box. Empty list if nothing was detected or the edgetpu node is not publishing.

Example:

detections = rc.vision.get_detections()
for det in detections:
    print(f"Found {det.class_id} at ({det.bbox[0]}, {det.bbox[1]})")

abstractmethod get_detections_async() → List[Detection]#: Returns the latest detections without the car in “go” mode.

Warning

This function breaks the start-update paradigm and should only be used in Jupyter Notebook.