1 VinOracleBatch

data_handling.VinOracleBatch(
    efm_snippet_view,
    candidate_poses_world_cam,
    reference_pose_world_rig,
    rri,
    pm_dist_before,
    pm_dist_after,
    pm_acc_before,
    pm_comp_before,
    pm_acc_after,
    pm_comp_after,
    scene_id,
    snippet_id,
    p3d_cameras,
    candidate_count=None,
    backbone_out=None,
    gt_obbs=None,
    detected_obbs=None,
    trajectory=None,
)

Single-snippet VIN training batch produced from an oracle label run.

1.1 Attributes

Name	Type	Description
efm_snippet_view	EfmSnippetView \| VinSnippetView \| None	EFM snippet view or minimal VIN snippet (None when loading from cache).
candidate_poses_world_cam	PoseTW	`PoseTW["N 12"]` or `PoseTW["B N 12"]` candidate poses as world←camera.
reference_pose_world_rig	PoseTW	`PoseTW["12"]` or `PoseTW["B 12"]` reference pose as world←rig_reference.
rri	Tensor	`Tensor["N", float32]` or `Tensor["B N", float32]` oracle RRI per candidate.
pm_dist_before	Tensor	`Tensor["N", float32]` or `Tensor["B N", float32]` Chamfer distance before (broadcasted).
pm_dist_after	Tensor	`Tensor["N", float32]` or `Tensor["B N", float32]` Chamfer distance after (per-candidate).
pm_acc_before	Tensor	`Tensor["N", float32]` or `Tensor["B N", float32]` accuracy distance before.
pm_comp_before	Tensor	`Tensor["N", float32]` or `Tensor["B N", float32]` completeness distance before.
pm_acc_after	Tensor	`Tensor["N", float32]` or `Tensor["B N", float32]` accuracy distance after.
pm_comp_after	Tensor	`Tensor["N", float32]` or `Tensor["B N", float32]` completeness distance after.
p3d_cameras	PerspectiveCameras	PyTorch3D cameras used for depth rendering/unprojection (same ordering as candidates).
candidate_count	Tensor \| None	Number of valid candidates (scalar or batched vector). When absent, runtime code falls back to the full candidate width.
scene_id	str \| list[str]	ASE scene id for diagnostics (string or list when batched).
snippet_id	str \| list[str]	Snippet id (tar key/url stem) for diagnostics (string or list when batched).
backbone_out	EvlBackboneOutput \| None	Optional cached EVL backbone outputs.

1.2 Methods

Name	Description
resolved_candidate_count	Return the valid candidate count as a scalar or batched vector.
candidate_valid_mask	Return a prefix mask that marks valid candidates.
shape_summary	Summarize tensor shapes for diagnostics/logging.
shuffle_candidates	Return a copy with candidate ordering randomly permuted.
from_label	Build a VIN oracle batch from an online label batch.
collate	Collate cached VIN batches by padding candidate sets to a shared length.

1.2.1 resolved_candidate_count

data_handling.VinOracleBatch.resolved_candidate_count(device=None)

Return the valid candidate count as a scalar or batched vector.

1.2.2 candidate_valid_mask

data_handling.VinOracleBatch.candidate_valid_mask(device=None)

Return a prefix mask that marks valid candidates.

1.2.3 shape_summary

data_handling.VinOracleBatch.shape_summary()

Summarize tensor shapes for diagnostics/logging.

1.2.4 shuffle_candidates

data_handling.VinOracleBatch.shuffle_candidates(generator=None)

Return a copy with candidate ordering randomly permuted.

The permutation is applied consistently across candidate-specific fields: poses, oracle RRI targets, per-metric distances, and the corresponding PyTorch3D cameras. Non-candidate fields (snippet view, backbone outputs) are preserved.

1.2.5 from_label

data_handling.VinOracleBatch.from_label(label_batch, *, efm_keep_keys)

Build a VIN oracle batch from an online label batch.

1.2.6 collate

data_handling.VinOracleBatch.collate(samples)

Collate cached VIN batches by padding candidate sets to a shared length.

# VinOracleBatch { #aria_nbv.data_handling.VinOracleBatch } ```python data_handling.VinOracleBatch( efm_snippet_view, candidate_poses_world_cam, reference_pose_world_rig, rri, pm_dist_before, pm_dist_after, pm_acc_before, pm_comp_before, pm_acc_after, pm_comp_after, scene_id, snippet_id, p3d_cameras, candidate_count=None, backbone_out=None, gt_obbs=None, detected_obbs=None, trajectory=None, ) ``` Single-snippet VIN training batch produced from an oracle label run. ## Attributes {.doc-section .doc-section-attributes} | Name | Type | Description | |---------------------------|------------------------------------------|--------------------------------------------------------------------------------------------------------------------------| | efm_snippet_view | EfmSnippetView \| VinSnippetView \| None | EFM snippet view or minimal VIN snippet (None when loading from cache). | | candidate_poses_world_cam | PoseTW | ``PoseTW["N 12"]`` or ``PoseTW["B N 12"]`` candidate poses as world←camera. | | reference_pose_world_rig | PoseTW | ``PoseTW["12"]`` or ``PoseTW["B 12"]`` reference pose as world←rig_reference. | | rri | Tensor | ``Tensor["N", float32]`` or ``Tensor["B N", float32]`` oracle RRI per candidate. | | pm_dist_before | Tensor | ``Tensor["N", float32]`` or ``Tensor["B N", float32]`` Chamfer distance before (broadcasted). | | pm_dist_after | Tensor | ``Tensor["N", float32]`` or ``Tensor["B N", float32]`` Chamfer distance after (per-candidate). | | pm_acc_before | Tensor | ``Tensor["N", float32]`` or ``Tensor["B N", float32]`` accuracy distance before. | | pm_comp_before | Tensor | ``Tensor["N", float32]`` or ``Tensor["B N", float32]`` completeness distance before. | | pm_acc_after | Tensor | ``Tensor["N", float32]`` or ``Tensor["B N", float32]`` accuracy distance after. | | pm_comp_after | Tensor | ``Tensor["N", float32]`` or ``Tensor["B N", float32]`` completeness distance after. | | p3d_cameras | PerspectiveCameras | PyTorch3D cameras used for depth rendering/unprojection (same ordering as candidates). | | candidate_count | Tensor \| None | Number of valid candidates (scalar or batched vector). When absent, runtime code falls back to the full candidate width. | | scene_id | str \| list\[str\] | ASE scene id for diagnostics (string or list when batched). | | snippet_id | str \| list\[str\] | Snippet id (tar key/url stem) for diagnostics (string or list when batched). | | backbone_out | EvlBackboneOutput \| None | Optional cached EVL backbone outputs. | ## Methods | Name | Description | | --- | --- | | [resolved_candidate_count](#aria_nbv.data_handling.VinOracleBatch.resolved_candidate_count) | Return the valid candidate count as a scalar or batched vector. | | [candidate_valid_mask](#aria_nbv.data_handling.VinOracleBatch.candidate_valid_mask) | Return a prefix mask that marks valid candidates. | | [shape_summary](#aria_nbv.data_handling.VinOracleBatch.shape_summary) | Summarize tensor shapes for diagnostics/logging. | | [shuffle_candidates](#aria_nbv.data_handling.VinOracleBatch.shuffle_candidates) | Return a copy with candidate ordering randomly permuted. | | [from_label](#aria_nbv.data_handling.VinOracleBatch.from_label) | Build a VIN oracle batch from an online label batch. | | [collate](#aria_nbv.data_handling.VinOracleBatch.collate) | Collate cached VIN batches by padding candidate sets to a shared length. | ### resolved_candidate_count { #aria_nbv.data_handling.VinOracleBatch.resolved_candidate_count } ```python data_handling.VinOracleBatch.resolved_candidate_count(device=None) ``` Return the valid candidate count as a scalar or batched vector. ### candidate_valid_mask { #aria_nbv.data_handling.VinOracleBatch.candidate_valid_mask } ```python data_handling.VinOracleBatch.candidate_valid_mask(device=None) ``` Return a prefix mask that marks valid candidates. ### shape_summary { #aria_nbv.data_handling.VinOracleBatch.shape_summary } ```python data_handling.VinOracleBatch.shape_summary() ``` Summarize tensor shapes for diagnostics/logging. ### shuffle_candidates { #aria_nbv.data_handling.VinOracleBatch.shuffle_candidates } ```python data_handling.VinOracleBatch.shuffle_candidates(generator=None) ``` Return a copy with candidate ordering randomly permuted. The permutation is applied consistently across candidate-specific fields: poses, oracle RRI targets, per-metric distances, and the corresponding PyTorch3D cameras. Non-candidate fields (snippet view, backbone outputs) are preserved. ### from_label { #aria_nbv.data_handling.VinOracleBatch.from_label } ```python data_handling.VinOracleBatch.from_label(label_batch, *, efm_keep_keys) ``` Build a VIN oracle batch from an online label batch. ### collate { #aria_nbv.data_handling.VinOracleBatch.collate } ```python data_handling.VinOracleBatch.collate(samples) ``` Collate cached VIN batches by padding candidate sets to a shared length.