Datasets:

louisye
/

social-robotics-saf

video_id stringlengths 36 36	source_dataset stringclasses 1 value	task_labels stringlengths 5 202	duration_sec float64 785 17.2k	fps float64 30 30	reasonable_emotion_tasks_analyzed_raw stringlengths 398 6.07k ⌀	reasonable_emotion_avg_task_score float64 -0.6 0.42 ⌀	reasonable_emotion_emotion_source stringclasses 1 value	acoustic_prosody_tasks_analyzed_raw stringlengths 4.79k 33.7k	acoustic_prosody_avg_prosody_scalar float64 -0.5 0.5	proxemic_kinematics_tasks_analyzed_raw stringlengths 3k 18.4k	proxemic_kinematics_max_proxemic_confidence float64 0 1	proxemic_kinematics_max_abs_proxemic_vector float64 -0.75 0.89	proxemic_kinematics_any_approach_detected bool 2 classes	proxemic_kinematics_any_avoidance_detected bool 2 classes	affirmation_gesture_tasks_analyzed_raw stringlengths 2 26.1k	affirmation_gesture_max_gesture_confidence float64 0 1 ⌀	affirmation_gesture_any_nod_detected bool 2 classes	affirmation_gesture_any_shake_detected bool 2 classes	affirmation_gesture_skipped_reason stringclasses 1 value	motor_resonance_tasks_analyzed_raw stringlengths 2.13k 13k	motor_resonance_max_ego_chaos_score float64 1 1	motor_resonance_max_empathy_scalar float64 0 1	motor_resonance_any_motor_resonance bool 2 classes	motor_resonance_max_mirroring_scalar float64 0 1	motor_resonance_any_mirroring bool 2 classes
004a1802-c546-4dcc-86ba-bf1080077017	ego4d	Grocery shopping indoors	1,446.8	30	null	null	null	[{"task_id": "t_01", "task_reaction_window_sec": [14.3, 16.3], "prosody_metrics": {"max_amplitude_dbFS": -3.375277587776353, "pitch_contour_variance": 0.003345294130538796, "emotion_scores": {"angry": 0.000704136211425066, "disgusted": 0.005960556212812662, "fearful": 0.0003219260252080858, "happy": 0.00102429825346916...	0.2	[{"task_id": "t_01", "per_person": [{"person_id": 220, "bbox_scale_delta_pct": 17.92, "depth_anything_v2_delta": 0.0, "proxemic_vector": 0.0, "classified_action": "Neutral", "proxemic_confidence": 0.0, "optical_flow_noise": 31.12, "measurement_window_sec": [927.0, 930.0], "window_source": "bystander_anchored", "min_con...	1	0.22	false	false	[{"task_id": "t_01", "segment_index": 0, "reaction_window_sec": [14.3, 16.3], "per_person": [{"person_id": 165, "pitch_oscillation_hz": 0.49, "yaw_oscillation_hz": 0.0, "interpolated_fraction": 0.173, "gesture_detected": "none", "confidence": 0.0, "measurement_window_sec": [733.0, 737.0], "window_source": "bystander_an...	0	false	false	null	[{"task_id": "t_01", "ego_kinetic_chaos_score": 1.0, "per_person": [{"person_id": 220, "bystander_pose_velocity_peak": 2.57, "resonance_delay_sec": 0.0, "motor_resonance_detected": false, "empathy_scalar": 0.0, "mirroring_detected": false, "mirroring_scalar": 0.0}, {"person_id": 62, "bystander_pose_velocity_peak": 0.96...	1	0	false	0	false
0219ad48-8f54-4f61-b22f-4d1e8173e584	ego4d	Cleaning / laundry, Making coffee	3,230.133333	30	null	null	null	[{"task_id": "t_01", "task_reaction_window_sec": [17.07, 19.07], "prosody_metrics": {"max_amplitude_dbFS": -12.95371961383789, "pitch_contour_variance": 0.22267204966579698, "emotion_scores": {"angry": 0.00011367886327207088, "disgusted": 0.0007337003480643034, "fearful": 0.00032152311177924275, "happy": 0.002292079152...	0.15	[{"task_id": "t_01", "per_person": [{"person_id": 208, "bbox_scale_delta_pct": -6.09, "depth_anything_v2_delta": 0.0, "proxemic_vector": 0.0, "classified_action": "Neutral", "proxemic_confidence": 0.0, "optical_flow_noise": 53.91, "measurement_window_sec": [2532.0, 2541.0], "window_source": "bystander_anchored", "min_c...	1	0.09	false	false	[{"task_id": "t_02", "segment_index": 3, "reaction_window_sec": [2234.4, 2236.4], "per_person": [{"person_id": 156, "pitch_oscillation_hz": 0.0, "yaw_oscillation_hz": 0.0, "interpolated_fraction": 0.055, "gesture_detected": "none", "confidence": 0.0, "measurement_window_sec": [2234.4, 2236.4], "window_source": "reactio...	0	false	false	null	[{"task_id": "t_01", "ego_kinetic_chaos_score": 1.0, "per_person": [{"person_id": 208, "bystander_pose_velocity_peak": 6.48, "resonance_delay_sec": 0.0, "motor_resonance_detected": false, "empathy_scalar": 0.0, "mirroring_detected": false, "mirroring_scalar": 0.0}, {"person_id": 156, "bystander_pose_velocity_peak": 4.4...	1	0	false	0.43	true
035eb249-9923-41ec-9f59-3b131c12bb1f	ego4d	Participating in a meeting, Talking to colleagues	2,074.333333	30	[{"task_id": "t_01", "task_label": "Participating in a meeting", "task_reaction_window_sec": [322.17, 324.17], "per_person": [{"person_id": 40, "temporal_slices": [{"slice_id": 1, "window_sec": [318.99, 319.32], "transition_pair": ["disgust", "disgust"], "terminal_magnitude": 0.42, "classified_direction": "neutral", "s...	0	hsemotion_onnx	[{"task_id": "t_01", "task_reaction_window_sec": [82.1, 84.1], "prosody_metrics": {"max_amplitude_dbFS": -13.443056215521063, "pitch_contour_variance": 0.009752398836795685, "emotion_scores": {"angry": 0.009724936448037624, "disgusted": 0.008021562360227108, "fearful": 0.0021855831146240234, "happy": 0.0179252848029136...	-0.025	[{"task_id": "t_01", "per_person": [{"person_id": 61, "bbox_scale_delta_pct": -1.24, "depth_anything_v2_delta": -0.0078, "proxemic_vector": 0.0, "classified_action": "Neutral", "proxemic_confidence": 0.0, "optical_flow_noise": 2.4, "measurement_window_sec": [447.0, 450.0], "window_source": "bystander_anchored", "min_co...	1	-0.14	false	false	[{"task_id": "t_01", "segment_index": 0, "reaction_window_sec": [82.1, 84.1], "per_person": [{"person_id": 36, "pitch_oscillation_hz": 0.0, "yaw_oscillation_hz": 0.0, "interpolated_fraction": 0.0, "gesture_detected": "none", "confidence": 0.0, "measurement_window_sec": [193.0, 200.0], "window_source": "bystander_anchor...	0	false	false	null	[{"task_id": "t_01", "ego_kinetic_chaos_score": 1.0, "per_person": [{"person_id": 61, "bystander_pose_velocity_peak": 1.06, "resonance_delay_sec": 0.0, "motor_resonance_detected": false, "empathy_scalar": 0.0, "mirroring_detected": false, "mirroring_scalar": 0.0}, {"person_id": 65, "bystander_pose_velocity_peak": 1.11,...	1	0.76	true	0	false
045451d6-2916-4c07-8e47-7cfdaa579086	ego4d	Attending a party, Playing board games	2,551.166667	30	[{"task_id": "t_01", "task_label": "Attending a party", "task_reaction_window_sec": [146.47, 148.47], "per_person": [{"person_id": 3, "temporal_slices": [{"slice_id": 1, "window_sec": [146.47, 147.46], "transition_pair": ["joy", "joy"], "terminal_magnitude": 0.4, "classified_direction": "positive", "slice_success_scala...	0.1144	hsemotion_onnx	[{"task_id": "t_01", "task_reaction_window_sec": [78.07, 80.07], "prosody_metrics": {"max_amplitude_dbFS": -11.575675508648812, "pitch_contour_variance": 0.031668576222519354, "emotion_scores": {"angry": 0.03308103233575821, "disgusted": 0.08761519938707352, "fearful": 0.010468565858900547, "happy": 0.49102723598480225...	0.05	[{"task_id": "t_01", "per_person": [{"person_id": 231, "bbox_scale_delta_pct": 69.13, "depth_anything_v2_delta": -0.0196, "proxemic_vector": 0.42, "classified_action": "Approach_Intervention", "proxemic_confidence": 1.0, "optical_flow_noise": 8.69, "measurement_window_sec": [1146.0, 1149.0], "window_source": "bystander...	1	0.46	true	true	[{"task_id": "t_01", "segment_index": 0, "reaction_window_sec": [78.07, 80.07], "per_person": [{"person_id": 2, "pitch_oscillation_hz": 0.0, "yaw_oscillation_hz": 0.0, "interpolated_fraction": 0.25, "gesture_detected": "none", "confidence": 0.0, "measurement_window_sec": [78.07, 80.07], "window_source": "reaction_windo...	0.8	true	true	null	[{"task_id": "t_01", "ego_kinetic_chaos_score": 1.0, "per_person": [{"person_id": 231, "bystander_pose_velocity_peak": 3.13, "resonance_delay_sec": 0.0, "motor_resonance_detected": false, "empathy_scalar": 0.0, "mirroring_detected": false, "mirroring_scalar": 0.0}, {"person_id": 153, "bystander_pose_velocity_peak": 3.3...	1	1	true	0.92	true
050e156a-fbad-45a7-8db5-2241cf7fb306	ego4d	Attending a TA session, Talking to colleagues, Talking with friends/housemates	1,902.3	30	"[{\"task_id\": \"t_02\", \"task_label\": \"Talking to colleagues\", \"task_reaction_window_sec\": [(...TRUNCATED)	0	hsemotion_onnx	"[{\"task_id\": \"t_01\", \"task_reaction_window_sec\": [55.63, 57.63], \"prosody_metrics\": {\"max_(...TRUNCATED)	-0.0625	"[{\"task_id\": \"t_01\", \"per_person\": [{\"person_id\": 90, \"bbox_scale_delta_pct\": -1.6, \"dep(...TRUNCATED)	1	-0.12	false	false	"[{\"task_id\": \"t_01\", \"segment_index\": 0, \"reaction_window_sec\": [55.63, 57.63], \"per_perso(...TRUNCATED)	0	false	false	null	"[{\"task_id\": \"t_01\", \"ego_kinetic_chaos_score\": 1.0, \"per_person\": [{\"person_id\": 90, \"b(...TRUNCATED)	1	1	true	0	false
0727c62a-a9c3-4783-9ea0-3499d7a7c987	ego4d	On a screen (phone/laptop), Tourism, Walking on street	6,033.033333	30	null	null	null	"[{\"task_id\": \"t_01\", \"task_reaction_window_sec\": [102.33, 104.33], \"prosody_metrics\": {\"ma(...TRUNCATED)	0	"[{\"task_id\": \"t_01\", \"per_person\": [{\"person_id\": 1966, \"bbox_scale_delta_pct\": 9.23, \"d(...TRUNCATED)	1	0.27	false	false	[]	null	null	null	mixed_skip	"[{\"task_id\": \"t_01\", \"ego_kinetic_chaos_score\": 1.0, \"per_person\": [{\"person_id\": 1966, \(...TRUNCATED)	1	0	false	0.88	true
0752c643-18c8-4fd3-9a32-7ec985f2a6bd	ego4d	Eating, On a screen (phone/laptop), Visiting exhibition, Walking on street	10,633.7	30	null	null	null	"[{\"task_id\": \"t_01\", \"task_reaction_window_sec\": [236.2, 238.2], \"prosody_metrics\": {\"max_(...TRUNCATED)	0	"[{\"task_id\": \"t_01\", \"per_person\": [{\"person_id\": 2153, \"bbox_scale_delta_pct\": -2.28, \"(...TRUNCATED)	1	-0.41	false	true	[]	null	null	null	mixed_skip	"[{\"task_id\": \"t_01\", \"ego_kinetic_chaos_score\": 1.0, \"per_person\": [{\"person_id\": 2153, \(...TRUNCATED)	1	0	false	0.53	true
0793bbe0-b8d5-4d46-9f02-c71d1bd4fad2	ego4d	"Cooking, Indoor Navigation (walking), Talking on the phone, Talking with family members, Watching t(...TRUNCATED)	3,507.566667	30	null	null	null	"[{\"task_id\": \"t_01\", \"task_reaction_window_sec\": [91.57, 93.57], \"prosody_metrics\": {\"max_(...TRUNCATED)	0.1061	"[{\"task_id\": \"t_01\", \"per_person\": [{\"person_id\": 46, \"bbox_scale_delta_pct\": 4.35, \"dep(...TRUNCATED)	1	0.4	true	false	[]	null	null	null	mixed_skip	"[{\"task_id\": \"t_01\", \"ego_kinetic_chaos_score\": 0.16, \"per_person\": [{\"person_id\": 46, \"(...TRUNCATED)	1	0	false	0	false
07b1c874-9dc1-42bc-87ff-dffa9bef14fb	ego4d	Baker	5,409.866667	30	null	null	null	"[{\"task_id\": \"t_01\", \"task_reaction_window_sec\": [97.7, 99.7], \"prosody_metrics\": {\"max_am(...TRUNCATED)	0	"[{\"task_id\": \"t_01\", \"per_person\": [{\"person_id\": 509, \"bbox_scale_delta_pct\": -45.83, \"(...TRUNCATED)	1	-0.34	false	true	[]	null	null	null	mixed_skip	"[{\"task_id\": \"t_01\", \"ego_kinetic_chaos_score\": 1.0, \"per_person\": [{\"person_id\": 509, \"(...TRUNCATED)	1	0	false	0	false
07d824bc-a3fd-4acd-8179-75a7e1e077ce	ego4d	Playing board games	1,800.133333	30	"[{\"task_id\": \"t_01\", \"task_label\": \"Playing board games\", \"task_reaction_window_sec\": [96(...TRUNCATED)	0	hsemotion_onnx	"[{\"task_id\": \"t_01\", \"task_reaction_window_sec\": [259.97, 261.97], \"prosody_metrics\": {\"ma(...TRUNCATED)	-0.05	"[{\"task_id\": \"t_01\", \"per_person\": [{\"person_id\": 115, \"bbox_scale_delta_pct\": 8.39, \"de(...TRUNCATED)	1	-0.2	false	false	"[{\"task_id\": \"t_01\", \"segment_index\": 0, \"reaction_window_sec\": [259.97, 261.97], \"per_per(...TRUNCATED)	0	false	false	null	"[{\"task_id\": \"t_01\", \"ego_kinetic_chaos_score\": 1.0, \"per_person\": [{\"person_id\": 115, \"(...TRUNCATED)	1	0	false	0	false

End of preview. Expand in Data Studio

Social Robotics — Social-Affective Filter (SAF), combined metadata

Dehydrated social-signal metadata from 200 egocentric Ego4D clips — the combined output of the Social-Affective Filter (SAF) layers in one wide table, so robots can learn to read human reactions. No raw video pixels and no audio. Each row is one source video keyed by video_id; rehydrate against your own legally-obtained Ego4D copies (below).

Each layer is also published standalone with a detailed card: attention (03a) · reasonable-emotion (03b) · acoustic-prosody (03c) · proxemic-kinematics (03d) · affirmation-gesture (03e) · motor-resonance (03f)

Columns (200 rows × 26 cols)

Manifest base

column	type	meaning
`video_id`	object	Ego4D source-clip UUID — the rehydration key (map to your own `<video_id>.mp4`).
`source_dataset`	object	Origin corpus (`ego4d`).
`task_labels`	object	Comma-joined VLM task label(s) — the activity the camera-wearer performed.
`duration_sec`	float64	Source clip duration (seconds).
`fps`	float64	Source clip frame rate.

03b · Reasonable Emotion — was the bystander's emotion appropriate?

column	type	meaning
`reasonable_emotion_avg_task_score`	float64	Mean per-task emotion-appropriateness scalar (−1 negative … +1 positive): was the bystander's facial-emotion trajectory contextually appropriate to the task?
`reasonable_emotion_emotion_source`	object	Facial-emotion backend used (`hsemotion_onnx`).

03c · Acoustic Prosody — ambient vocal tone (corroboration only)

column	type	meaning
`acoustic_prosody_avg_prosody_scalar`	float64	Mean ambient acoustic tone (−1 alarming … +1 soothing). Corroborating context, not bystander-attributed (whole-scene audio).

03d · Proxemic Kinematics — approach vs avoidance

column	type	meaning
`proxemic_kinematics_max_proxemic_confidence`	float64	Max per-bystander proxemic confidence (sign agreement of bbox-scale vs depth delta).
`proxemic_kinematics_max_abs_proxemic_vector`	float64	Strongest approach/avoidance magnitude in the clip (−1 hard recoil … +1 hard lunge).
`proxemic_kinematics_any_approach_detected`	bool	True if any bystander approached / intervened.
`proxemic_kinematics_any_avoidance_detected`	bool	True if any bystander recoiled / avoided.

03e · Affirmation Gesture — nod vs shake (head-pose only)

column	type	meaning
`affirmation_gesture_max_gesture_confidence`	float64	Max head-gesture confidence. Head-pose only (gaze proven noise and discarded).
`affirmation_gesture_any_nod_detected`	object	True if any bystander gave an affirming nod.
`affirmation_gesture_any_shake_detected`	object	True if any bystander gave a negating head-shake.
`affirmation_gesture_skipped_reason`	object	Why a clip produced no trustworthy gesture (e.g. `no_head_pose`).

03f · Motor Resonance — sympathetic flinch / mirroring

column	type	meaning
`motor_resonance_max_ego_chaos_score`	float64	Max wearer ego-motion chaos (camera-jolt severity, the stimulus).
`motor_resonance_max_empathy_scalar`	float64	Max sympathetic-flinch empathy scalar.
`motor_resonance_any_motor_resonance`	bool	True if any bystander flinched in sympathy with a camera jolt.
`motor_resonance_max_mirroring_scalar`	float64	Max spine-angle mirroring congruence.
`motor_resonance_any_mirroring`	bool	True if any bystander mirrored the wearer's motion.

Per-layer raw detail

Each layer also carries a *_raw column — the full per-task / per-person nested detail as a JSON string (for research drill-down): reasonable_emotion_tasks_analyzed_raw, acoustic_prosody_tasks_analyzed_raw, proxemic_kinematics_tasks_analyzed_raw, affirmation_gesture_tasks_analyzed_raw, motor_resonance_tasks_analyzed_raw.

How to use

Download social_metadata.parquet (Pandas/HF datasets ready).
Use rehydrate_dataset.py with your legally-obtained Ego4D copies to map features back to footage by video_id.

Provenance

export_metadata.json carries schema_version (additive-only + column-hash), export_timestamp, active_layers, and pipeline_git_sha.

Downloads last month: 37