Interface: ObjectDetectionOptions
Defined in: src/world/objects/ObjectDetector.ts:52
Inputs for one object detection run, without changing shared options.
Properties
backend?
optionalbackend:"gemini"|"mediapipe"
Defined in: src/world/objects/ObjectDetector.ts:54
Uses this backend for this run only; otherwise uses the configured backend.
snapshot?
optionalsnapshot:CameraSnapshot
Defined in: src/world/objects/ObjectDetector.ts:61
Uses this frame instead of capturing another one. Gemini requires
base64; MediaPipe requires imageData. Both may be supplied for the
same frame. Capture and submit in the same frame so the camera pose and
depth mesh match, and do not modify the pixels until the run completes.