Detection

ark_load_detection_model(model_path) => error code

Load an existing detection model on the AI Accelator.

ark_set_detection_params(min_confidence, labels=[], context=”on_surface”, context_model=”table”) => void

Configure parameters for the detection including confidence thresholds and labels of objects to be returned. The context parameter defines how 3D pose estimation is performed.

• If context == “on_surface” (default): 3d pose of a detected object is inferred from the 2d location of the object in the image, geometric information of the camera and the constraint that the object lie on a surface (specified by the context_model variable, set to “table” by default)

• If context == “fp”: 3d pose is estimated by FoundationPose. Note that in this case, the context_model parameter is not used

ark_infer_detection() => error code, 3d poses, frame_ids, scores

Request a detection from the model, returns any relevant error codes, the 3D poses of the detected objects and the confidence of each detection, along with a frame_id to be passed to subsequent processes.