jianyong365/objgauss-nerf-lego-near1m-model
Updated
Dataset Viewer
The dataset viewer is not available for this split.
Cannot load the dataset split (in streaming mode) to extract the first rows.
Error code: StreamingRowsError
Exception: CastError
Message: Couldn't cast
asset_id: string
dataset: string
input: string
input_format: string
gaussian_source: string
note: string
gaussian_count: int64
fields: list<item: string>
child 0, item: string
gaussian_ply: string
splat_path: string
mask_manifest: string
slots: int64
initial_field: string
trained_field: string
object_ply: string
object_label_policy: null
background_training: null
public_splat: null
public_object_ply: null
training: struct<initial_loss: double, final_loss: double, iterations: int64, supervised_gaussians: int64, fra (... 795 chars omitted)
child 0, initial_loss: double
child 1, final_loss: double
child 2, iterations: int64
child 3, supervised_gaussians: int64
child 4, frames: int64
child 5, projected: int64
child 6, matched: int64
child 7, vote_quality: struct<gaussian_count: int64, slots: int64, supervised_gaussians: int64, unsupervised_gaussians: int (... 514 chars omitted)
child 0, gaussian_count: int64
child 1, slots: int64
child 2, supervised_gaussians: int64
child 3, unsupervised_gaussians: int64
child 4, supervised_fraction: double
child 5, projected: int64
child 6, matched: int64
child 7, matched_projected_fraction: double
child 8, observation_weight: struct<min: double, mean: double, max: double>
child 0, min: double
child 1, mean: double
child 2, max: double
child 9, vote_conflict: struct<gaussians: int64, fraction: double, target_entropy: double, normalized_tar
...
string, device: string, points_per_side: int64, pred_iou_thre (... 115 chars omitted)
child 0, model_type: string
child 1, checkpoint: string
child 2, device: string
child 3, points_per_side: int64
child 4, pred_iou_thresh: double
child 5, stability_score_thresh: double
child 6, min_area: int64
child 7, max_area_fraction: double
child 8, max_masks_per_frame: int64
source: string
width: int64
source_type: string
split: string
frames: list<item: struct<name: string, image_path: string, transform_matrix: list<item: list<item: double>> (... 135 chars omitted)
child 0, item: struct<name: string, image_path: string, transform_matrix: list<item: list<item: double>>, masks: li (... 123 chars omitted)
child 0, name: string
child 1, image_path: string
child 2, transform_matrix: list<item: list<item: double>>
child 0, item: list<item: double>
child 0, item: double
child 3, masks: list<item: struct<slot: int64, label: string, mask_path: string, confidence: double, area: int64, bb (... 24 chars omitted)
child 0, item: struct<slot: int64, label: string, mask_path: string, confidence: double, area: int64, bbox: list<it (... 12 chars omitted)
child 0, slot: int64
child 1, label: string
child 2, mask_path: string
child 3, confidence: double
child 4, area: int64
child 5, bbox: list<item: double>
child 0, item: double
to
{'width': Value('int64'), 'height': Value('int64'), 'camera_angle_x': Value('float64'), 'source': Value('string'), 'source_type': Value('string'), 'split': Value('string'), 'slots': List({'slot': Value('int64'), 'label': Value('string')}), 'sam': {'model_type': Value('string'), 'checkpoint': Value('string'), 'device': Value('string'), 'points_per_side': Value('int64'), 'pred_iou_thresh': Value('float64'), 'stability_score_thresh': Value('float64'), 'min_area': Value('int64'), 'max_area_fraction': Value('float64'), 'max_masks_per_frame': Value('int64')}, 'frames': List({'name': Value('string'), 'image_path': Value('string'), 'transform_matrix': List(List(Value('float64'))), 'masks': List({'slot': Value('int64'), 'label': Value('string'), 'mask_path': Value('string'), 'confidence': Value('float64'), 'area': Value('int64'), 'bbox': List(Value('float64'))})}), 'split_manifest': {'source': Value('string'), 'kind': Value('string'), 'heldout_every': Value('int64'), 'heldout_offset': Value('int64'), 'source_frames': Value('int64'), 'frames': Value('int64')}}
because column names don't match
Traceback: Traceback (most recent call last):
File "/src/services/worker/src/worker/utils.py", line 147, in get_rows_or_raise
return get_rows(
dataset=dataset,
...<4 lines>...
column_names=column_names,
)
File "/src/libs/libcommon/src/libcommon/utils.py", line 272, in decorator
return func(*args, **kwargs)
File "/src/services/worker/src/worker/utils.py", line 127, in get_rows
rows_plus_one = list(itertools.islice(safe_iter(ds, dataset=dataset), rows_max_number + 1))
File "/src/services/worker/src/worker/utils.py", line 478, in safe_iter
yield from ds.decode(False) if ds.features else ds
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2818, in __iter__
for key, example in ex_iterable:
^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2355, in __iter__
for key, pa_table in self._iter_arrow():
~~~~~~~~~~~~~~~~^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2380, in _iter_arrow
for key, pa_table in self.ex_iterable._iter_arrow():
~~~~~~~~~~~~~~~~~~~~~~~~~~~~^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 536, in _iter_arrow
for key, pa_table in iterator:
^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 419, in _iter_arrow
for key, pa_table in self.generate_tables_fn(**gen_kwags):
~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 343, in _generate_tables
self._cast_table(pa_table, json_field_paths=json_field_paths),
~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 132, in _cast_table
pa_table = table_cast(pa_table, self.info.features.arrow_schema)
File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2369, in table_cast
return cast_table_to_schema(table, schema)
File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2297, in cast_table_to_schema
raise CastError(
...<3 lines>...
)
datasets.table.CastError: Couldn't cast
asset_id: string
dataset: string
input: string
input_format: string
gaussian_source: string
note: string
gaussian_count: int64
fields: list<item: string>
child 0, item: string
gaussian_ply: string
splat_path: string
mask_manifest: string
slots: int64
initial_field: string
trained_field: string
object_ply: string
object_label_policy: null
background_training: null
public_splat: null
public_object_ply: null
training: struct<initial_loss: double, final_loss: double, iterations: int64, supervised_gaussians: int64, fra (... 795 chars omitted)
child 0, initial_loss: double
child 1, final_loss: double
child 2, iterations: int64
child 3, supervised_gaussians: int64
child 4, frames: int64
child 5, projected: int64
child 6, matched: int64
child 7, vote_quality: struct<gaussian_count: int64, slots: int64, supervised_gaussians: int64, unsupervised_gaussians: int (... 514 chars omitted)
child 0, gaussian_count: int64
child 1, slots: int64
child 2, supervised_gaussians: int64
child 3, unsupervised_gaussians: int64
child 4, supervised_fraction: double
child 5, projected: int64
child 6, matched: int64
child 7, matched_projected_fraction: double
child 8, observation_weight: struct<min: double, mean: double, max: double>
child 0, min: double
child 1, mean: double
child 2, max: double
child 9, vote_conflict: struct<gaussians: int64, fraction: double, target_entropy: double, normalized_tar
...
string, device: string, points_per_side: int64, pred_iou_thre (... 115 chars omitted)
child 0, model_type: string
child 1, checkpoint: string
child 2, device: string
child 3, points_per_side: int64
child 4, pred_iou_thresh: double
child 5, stability_score_thresh: double
child 6, min_area: int64
child 7, max_area_fraction: double
child 8, max_masks_per_frame: int64
source: string
width: int64
source_type: string
split: string
frames: list<item: struct<name: string, image_path: string, transform_matrix: list<item: list<item: double>> (... 135 chars omitted)
child 0, item: struct<name: string, image_path: string, transform_matrix: list<item: list<item: double>>, masks: li (... 123 chars omitted)
child 0, name: string
child 1, image_path: string
child 2, transform_matrix: list<item: list<item: double>>
child 0, item: list<item: double>
child 0, item: double
child 3, masks: list<item: struct<slot: int64, label: string, mask_path: string, confidence: double, area: int64, bb (... 24 chars omitted)
child 0, item: struct<slot: int64, label: string, mask_path: string, confidence: double, area: int64, bbox: list<it (... 12 chars omitted)
child 0, slot: int64
child 1, label: string
child 2, mask_path: string
child 3, confidence: double
child 4, area: int64
child 5, bbox: list<item: double>
child 0, item: double
to
{'width': Value('int64'), 'height': Value('int64'), 'camera_angle_x': Value('float64'), 'source': Value('string'), 'source_type': Value('string'), 'split': Value('string'), 'slots': List({'slot': Value('int64'), 'label': Value('string')}), 'sam': {'model_type': Value('string'), 'checkpoint': Value('string'), 'device': Value('string'), 'points_per_side': Value('int64'), 'pred_iou_thresh': Value('float64'), 'stability_score_thresh': Value('float64'), 'min_area': Value('int64'), 'max_area_fraction': Value('float64'), 'max_masks_per_frame': Value('int64')}, 'frames': List({'name': Value('string'), 'image_path': Value('string'), 'transform_matrix': List(List(Value('float64'))), 'masks': List({'slot': Value('int64'), 'label': Value('string'), 'mask_path': Value('string'), 'confidence': Value('float64'), 'area': Value('int64'), 'bbox': List(Value('float64'))})}), 'split_manifest': {'source': Value('string'), 'kind': Value('string'), 'heldout_every': Value('int64'), 'heldout_offset': Value('int64'), 'source_frames': Value('int64'), 'frames': Value('int64')}}
because column names don't matchNeed help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.
Development-stage release / 开发阶段发布。 This Hugging Face repository is currently used for ObjGauss development, verification, and reproducibility handoff. Assets, file layout, metrics, model weights, and documentation may change before a stable release.
ObjGauss NeRF Lego near-1M Trained Object-Aware Gaussian Asset
This dataset contains a local ObjGauss research artifact built from the NeRF Synthetic Lego scene with Nerfstudio Splatfacto and ObjGauss Object Field mask-vote training.
It is intended for open research download and reproducibility, not for commercial demonstration. The upstream scene comes from the NeRF Synthetic Lego example data, so downstream use must respect the upstream dataset terms.
Contents
gaussians/gaussians.splat: compact Spark/Splat viewer asset for quick visual loading.gaussians/object_aware_gaussians.ply: object-aware Gaussian PLY withobject_id.object_field/object_field_initial.npz: initialized Object Field logits.object_field/object_field_trained.npz: SAM mask-vote trained Object Field logits.masks/: SAM mask manifests and.npymasks used for Object Field supervision.manifests/training-output-manifest.json: ObjGauss registration manifest.manifests/mask-training-summary.json: mask-vote training summary.metrics/: emergence metrics, curve outputs, and current WebGPU production SLA summary.checksums.sha256: SHA256 checksums for uploaded release files.
The intentionally omitted file is gaussians.ply without object_id, because it
duplicates the 1.1GB object-aware PLY for this release. The full Nerfstudio
checkpoint is published in the companion model repository.
Key Facts
- Scene: NeRF Synthetic Lego
- Trainer: Nerfstudio Splatfacto
- ObjGauss asset id:
nerf-lego-splatfacto-near1m-random1300k-v1-candidate - Exported Gaussian count:
4,503,634 - Object-aware Gaussian count:
4,503,634 - Object Field slots:
4 - Supervised Gaussians:
632,522 - Mask-vote final loss:
0.069276 - Object emergence score:
0.846996
Usage
For quick visual inspection, load:
gaussians/gaussians.splat
For ObjGauss object editing and analysis, use:
gaussians/object_aware_gaussians.ply
object_field/object_field_trained.npz
manifests/training-output-manifest.json
ObjGauss local viewer support in this release uses .splat for fast first view
and delays loading the large object-aware PLY until object editing is requested.
Current Limitations
- This is a research artifact derived from NeRF Synthetic Lego.
- Object labels are produced by ObjGauss Object Field training with SAM mask voting, not by human semantic annotation.
- The near-1M scale gate passed, but the current
audit:webgpu-cpath-production-slasummary is included as failed. The failure is an audit/runtime path issue in the WebGPU presentation transition check, not absence of the trained object-aware PLY. - The generated object assignment should be treated as diagnostic/research evidence, not a production-ready semantic segmentation benchmark.
Companion Repositories
- Model checkpoint:
jianyong365/objgauss-nerf-lego-near1m-model - Source code:
https://github.com/sessgraph/ObjGauss
- Downloads last month
- 24
Models trained or fine-tuned on jianyong365/objgauss-nerf-lego-near1m