GRID Docs | General Robotics

Reconstruct camera-to-world poses + clip-relative depth from an ordered clip of frames.

Parameters

Sequence[ImageInput]

required

Ordered clip of 1-32 RGB frames (path, URL, PIL Image, or numpy array). Each call is a self-contained window; chunk longer sequences into overlapping clips and align them client-side.

float | None

Optional timeout in seconds for the HTTP request.

Returns

Dict[str, np.ndarray]: Dict with "extrinsic" (N,3,4) camera-to-world in OpenCV axes, "intrinsic" (N,3,3), "depth" (N,H,W), and "depth_conf" (N,H,W), all float32. Depth and translation are in one clip-relative scale (normalised by the clip’s mean anchor distance) — not metres; recover physical scale externally.

Example

from grid_cortex_client import CortexClient, ModelType
client = CortexClient()
result = client.run(ModelType.LINGBOT_MAP, images=frames)
print(result["extrinsic"].shape, result["depth"].shape)  # (N, 3, 4) (N, H, W)

Overview

Cortex Models

lingbot-map

Parameters

Returns

Example

​Parameters

​Returns

​Example

Parameters

Returns

Example