Aria Everyday Activities Dataset
Paper • 2402.13349 • Published • 31
Dataset Viewer
The dataset viewer is not available for this dataset.
Cannot get the config names for the dataset.
Error code: ConfigNamesError
Exception: TypeError
Message: 'str' object is not a mapping
Traceback: Traceback (most recent call last):
File "/src/services/worker/src/worker/job_runners/dataset/config_names.py", line 67, in compute_config_names_response
config_names = get_dataset_config_names(
path=dataset,
token=hf_token,
)
File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 161, in get_dataset_config_names
dataset_module = dataset_module_factory(
path,
...<4 lines>...
**download_kwargs,
)
File "/usr/local/lib/python3.14/site-packages/datasets/load.py", line 1182, in dataset_module_factory
raise e1 from None
File "/usr/local/lib/python3.14/site-packages/datasets/load.py", line 1157, in dataset_module_factory
).get_module()
~~~~~~~~~~^^
File "/usr/local/lib/python3.14/site-packages/datasets/load.py", line 614, in get_module
dataset_infos = DatasetInfosDict.from_dataset_card_data(dataset_card_data)
File "/usr/local/lib/python3.14/site-packages/datasets/info.py", line 399, in from_dataset_card_data
dataset_info = DatasetInfo._from_yaml_dict(dataset_card_data["dataset_info"])
File "/usr/local/lib/python3.14/site-packages/datasets/info.py", line 317, in _from_yaml_dict
yaml_data["features"] = Features._from_yaml_list(yaml_data["features"])
~~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/features/features.py", line 2178, in _from_yaml_list
return cls.from_dict(from_yaml_inner(yaml_data))
~~~~~~~~~~~~~~~^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/features/features.py", line 2174, in from_yaml_inner
return {name: from_yaml_inner(_feature) for name, _feature in zip(names, obj)}
~~~~~~~~~~~~~~~^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/features/features.py", line 2171, in from_yaml_inner
return {"_type": snakecase_to_camelcase(_type), **unsimplify(obj)[_type]}
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
TypeError: 'str' object is not a mappingNeed help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.
VEGAS: Human-Aligned Video Caption Evaluation via Gaze
VEGAS (Video caption Evaluation via GAze Score) is a multimodal benchmark designed to study gaze-conditioned video captioning and evaluation. It leverages human gaze as a direct proxy for visual attention to generate personalized, attention-aligned captions.
The dataset includes synchronized gaze trajectories, fixations, heatmaps, and multiple human-written captions from two distinct domains:
- SlideVQA: Presentation-centric static slide viewing with webcam-based gaze tracking.
- Aria Everyday Activities (AEA): Egocentric dynamic daily activities featuring wearers' physical gaze tracks recorded via Project Aria glasses.
Directory Structure
vegas_dataset_hf/
├── AriaEverydayActivities/
│ ├── aggregated_captions.csv
│ └── action_segments/
└── SlideVQA/
├── aggregated_captions.csv
├── metadata.csv
├── quality.csv
├── gaze/
├── fixation/
└── heatmap/
1. Slide Deck Captioning Dataset (SlideVQA)
Constructed using slide presentations from the SlideVQA dataset, featuring webcam eye-tracking via the RealEye platform.
- Stimuli & Participants: 30 slide decks (first 10 slides each), with 5 human annotators recruited via Prolific per deck.
- Protocol: Calibration, natural reading ($\ge$ 20s per slide), followed by writing a takeaway summary (30-200 chars).
Files
aggregated_captions.csv: Slide captions (columns:slide_id,slide_num,human_0tohuman_4,gemini_groundtruth).metadata.csv: Anonymized Prolific participant ID mapping.quality.csv: RealEye calibration quality score (1-6) per participant.gaze/: Raw millisecond-level gaze log CSVs (columns:participant_id,gaze_x_percents,gaze_y_percents,gaze_timestamp_ms, etc.).fixation/: Discrete eye fixation event CSVs (columns:fixation_point_x,fixation_point_y,fixation_duration_ms,item_cdn_url, etc.).heatmap/: Processed aggregation PNGs under<slide_id>/Anonymous_ID_<participant_id>/<slide_num>_heatmap.png.
2. Egocentric Video Captioning Dataset (Aria Everyday Activities)
Focuses on egocentric daily living activities utilizing recordings from the Aria Everyday Activities (AEA) dataset with wearer eye-tracking.
- Curation: Egocentric video streams segmented using
Gemini-3.1-Pro-Preview. Selected the top 25% gaze-critical clips (lowest SBERT similarity and highest VEGAS score evaluated byGemma-4-31B-IT), resulting in 332 segments. - Human Annotation: Prolific annotators viewed egocentric clips with wearers' gaze overlays and wrote first-person action descriptions ($\ge 30$ chars).
Files
aggregated_captions.csv: Segment captions (columns:vid,action_id,start_time,end_time,gemini_caption,human_0_captiontohuman_4_caption).action_segments/: JSON files detailing actions and time windows parsed byGemini-3.1-Pro-Preview.
Ethical and Privacy Considerations
- Visual Data Protection: webcam streams are processed locally; only absolute coordinate pairs are stored.
- De-identification: Prolific IDs are removed and replaced with random UUIDs.
- IRB Approval: All user studies and datasets have active Institutional Review Board approval.
- Downloads last month
- 3