1 vin.types.diagnostics

vin.types.diagnostics

Diagnostics payloads for active and legacy VIN scorer variants.

These dataclasses are canonical type owners even when the model producing them is still a deprecated experimental scorer. App panels and diagnostics helpers should import them through aria_nbv.vin.types or this leaf module, not through aria_nbv.vin.experimental.

1.1 Classes

Name Description
VinForwardDiagnostics Intermediate tensors produced during a VIN forward pass.
VinV3ForwardDiagnostics Candidate-aligned intermediate tensors from VIN-Core forward passes.

1.1.1 VinForwardDiagnostics

vin.types.diagnostics.VinForwardDiagnostics(
    backbone_out,
    candidate_center_rig_m,
    candidate_radius_m,
    candidate_center_dir_rig,
    candidate_forward_dir_rig,
    view_alignment,
    pose_enc,
    pose_vec,
    voxel_center_rig_m,
    voxel_radius_m,
    voxel_center_dir_rig,
    voxel_forward_dir_rig,
    voxel_view_alignment,
    voxel_pose_enc,
    voxel_pose_vec,
    field_in,
    field,
    global_feat,
    local_feat,
    tokens,
    token_valid,
    candidate_valid,
    voxel_valid_frac,
    feats,
)

Intermediate tensors produced during a VIN forward pass.

1.1.1.1 Attributes

Name Description
backbone_out EVL backbone outputs used to build the scene field.
candidate_center_rig_m Tensor["B N_q 3", float32] candidate centers in reference-rig metres.
candidate_radius_m Tensor["B N_q 1", float32] candidate radii in reference-rig metres.
candidate_center_dir_rig Tensor["B N_q 3", float32] unit center directions in the reference rig frame.
candidate_forward_dir_rig Tensor["B N_q 3", float32] unit forward directions in the reference rig frame.
view_alignment Tensor["B N_q 1", float32] forward-to-inward alignment in [-1, 1].
pose_enc Tensor["B N_q E_pose", float32] pose encoder output.
pose_vec Tensor["B N_q D_pose", float32] pose vector fed into LFF, if present.
voxel_center_rig_m Tensor["B 3", float32] Voxel-grid origin in the reference rig frame (or None).
voxel_radius_m Tensor["B 1", float32] Radius of the voxel-grid origin in the reference rig frame (or None).
voxel_center_dir_rig Tensor["B 3", float32] Unit direction to voxel origin in the reference rig frame (or None).
voxel_forward_dir_rig Tensor["B 3", float32] Voxel frame +Z axis in the reference rig frame (or None).
voxel_view_alignment Tensor["B 1", float32] Dot product of voxel forward and -center directions (or None).
voxel_pose_enc Tensor["B E_pose", float32] LFF-encoded voxel pose in the reference rig frame (or None).
voxel_pose_vec Tensor["B D_pose", float32] Voxel pose vector fed into LFF (or None).
field_in Tensor["B C_in D H W", float32] Raw scene field before projection.
field Tensor["B C_out D H W", float32] Projected scene field.
global_feat Tensor["B N_q C_global", float32] global candidate features, if present.
local_feat Tensor["B N_q C_out", float32] pooled frustum features.
tokens Tensor["B N_q K C_out", float32] sampled frustum tokens.
token_valid Tensor["B N_q K", bool] token validity mask.
candidate_valid Tensor["B N_q", bool] candidate validity diagnostic.
voxel_valid_frac Tensor["B N_q 1", float32] valid voxel/frustum fraction per candidate.
feats Tensor["B N_q F", float32] concatenated VIN features after masking.

1.1.2 VinV3ForwardDiagnostics

vin.types.diagnostics.VinV3ForwardDiagnostics(
    backbone_out,
    candidate_center_rig_m,
    pose_enc,
    pose_vec,
    field_in,
    field,
    global_feat,
    candidate_valid,
    feats,
    voxel_valid_frac=None,
    semidense_candidate_vis_frac=None,
    pos_grid=None,
    semidense_proj=None,
    semidense_grid_feat=None,
    voxel_proj=None,
    traj_feat=None,
    traj_ctx=None,
    traj_pose_vec=None,
    traj_pose_enc=None,
)

Candidate-aligned intermediate tensors from VIN-Core forward passes.

1.1.2.1 Attributes

Name Description
backbone_out EVL backbone outputs used to build the scene field.
candidate_center_rig_m Tensor["B N_q 3", float32] candidate centers in reference-rig metres.
pose_enc Tensor["B N_q E_pose", float32] pose encoder output.
pose_vec Tensor["B N_q D_pose", float32] pose vector fed into the encoder.
field_in Tensor["B C_in D H W", float32] Raw scene field before projection.
field Tensor["B C_out D H W", float32] Projected scene field.
global_feat Tensor["B N_q C_global", float32] pose-conditioned global features.
candidate_valid Tensor["B N_q", bool] diagnostic candidate-valid mask.
feats Tensor["B N_q F", float32] concatenated VIN candidate features.
voxel_valid_frac Tensor["B N_q", float32] candidate voxel-coverage proxy, if computed.
semidense_candidate_vis_frac Tensor["B N_q", float32] candidate semidense visibility, if computed.
semidense_valid_frac Read-only compatibility alias for semidense_candidate_vis_frac.
pos_grid Tensor["B 3 D H W", float32] Normalized voxel position grid (if computed).
semidense_proj Tensor["B N_q C_proj", float32] semidense projection features.
semidense_grid_feat Tensor["B N_q C_cnn", float32] CNN projection-grid features.
voxel_proj Tensor["B N_q C_proj", float32] voxel-projection features.
traj_feat Tensor["B F_traj", float32] Pooled trajectory embedding (if enabled).
traj_ctx Tensor["B N_q F_pose", float32] trajectory context per candidate query.
traj_pose_vec Tensor["B T D_v", float32] Per-frame trajectory pose vectors.
traj_pose_enc Tensor["B T F_traj", float32] Per-frame trajectory pose encodings.
Back to top