1 VinSnippetView

data_handling.efm_views.VinSnippetView(points_world, lengths, t_world_rig)

Minimal snippet payload for VIN v2 batching.

1.1 Attributes

Name Type Description
points_world Tensor Tensor["K (3+C)", float32] collapsed semidense points. Base columns are XYZ; optional extras include inv_dist_std and observation count (number of snippet frames that saw the point).
lengths Tensor Tensor["B"] or Tensor["1"] number of valid points in points_world (before padding).
t_world_rig PoseTW PoseTW["F 12"] historical world←rig poses.

1.2 Methods

Name Description
to Move the VIN snippet tensors to the requested device and dtype.

1.2.1 to

data_handling.efm_views.VinSnippetView.to(device, *, dtype=None)

Move the VIN snippet tensors to the requested device and dtype.