DepthPredictionResult

spatialhub.structures.DepthPredictionResult encapsulates estimated metric or relative depth maps and camera parameters produced by depth estimation adapters such as DepthAnything3Adapter.

from spatialhub.structures import DepthPredictionResult

---

Fields

Field	Type	Shape	Description
image	np.ndarray	(N, H, W, 3) or (H, W, 3) uint8	Input RGB image array(s).
depth	np.ndarray	(N, H, W) float32	Predicted depth map array in meters or relative scale.
conf	np.ndarray | None	(N, H, W) float32	Per-pixel prediction confidence, if provided by the model.
intrinsics	np.ndarray | None	(N, 3, 3) float32	Estimated camera intrinsic matrices [[fx, 0, cx], [0, fy, cy], [0, 0, 1]].
extrinsics	np.ndarray | None	(N, 4, 4) float32	Estimated camera extrinsic transformation matrices [R | t].
depth_type	str	"metric"	Depth interpretation scale: "metric", "relative", "inverse", or "disparity".

---

Usage Example

from spatialhub import DepthAnything3
from spatialhub.utils import reproject_depth_to_3d

estimator = DepthAnything3(model_variant="da3_metric_large")
result = estimator.predict("scene.png")

print(f"Depth shape: {result.depth.shape}")
print(f"Depth range: {result.depth.min():.2f}m to {result.depth.max():.2f}m")

# Reproject depth map to 3D point cloud using camera intrinsics
if result.intrinsics is not None:
    xyz_map = reproject_depth_to_3d(result.depth[0], result.intrinsics[0])