1 VinV3ForwardDiagnostics

vin.types.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,
    semidense_valid_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,
)

Diagnostics for VIN v3 (VIN-Core).

1.1 Attributes

Name	Description
backbone_out	EVL backbone outputs used to build the scene field.
candidate_center_rig_m	Tensor["B N 3", float32] Candidate centers in the reference rig frame.
pose_enc	Tensor["B N E_pose", float32] Pose encoder output.
pose_vec	Tensor["B N D_pose", float32] Pose vector fed into the pose 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 C_global", float32] Pose-conditioned global features.
candidate_valid	Tensor["B N", bool] Candidate validity mask.
feats	Tensor["B N F", float32] Concatenated VIN features.
voxel_valid_frac	Tensor["B N", float32] Per-candidate voxel coverage proxy (if computed).
semidense_candidate_vis_frac	Tensor["B N", float32] Per-candidate semidense visibility proxy (if computed).
semidense_valid_frac	Deprecated 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 C_proj", float32] Per-candidate semidense projection features.
semidense_grid_feat	Tensor["B N C_cnn", float32] CNN-encoded semidense projection grid features.
voxel_proj	Tensor["B N C_proj", float32] Per-candidate voxel projection features.
traj_feat	Tensor["B F_traj", float32] Pooled trajectory embedding (if enabled).
traj_ctx	Tensor["B N F_pose", float32] Trajectory context attended by pose tokens.
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.