id string | source string | subject_id string | session_id string | device_id string | game_id string | sensor_loc string | native_hz float64 | event_marker_available bool | telemetry_event string | event_onset_s float64 | panel_order list | panel_metrics dict | system_prompt string | target dict | assistant_text string | schema_version string | motion_state string | relative_motion_intensity string | temporal_pattern string | motor_strategy_hypothesis string | teacher_confidence float64 | training_value string | use_for_training bool | teacher_metadata dict | panel_images images list |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
mhealth_representation_only:mHealth_subject10:mHealth_subject10:mhealth_ankle:0 | mhealth_representation_only | mHealth_subject10 | mHealth_subject10 | mhealth_imu | non_game_representation_source | mhealth_ankle | 50 | false | no_event_marker | null | [
"p1_imu_attn",
"p2_accel_jerk",
"p3_burst_decomp",
"p4_energy_env",
"p5_jerk_timeline",
"p6_gyro",
"p7_spectrum",
"p8_spectrogram",
"p9_importance",
"p10_head_dispersion",
"p11_subject_reference_distance",
"p12_motion_tertile_similarity"
] | {
"n_bursts": 3,
"peak_jerk": 203.90803254201663,
"acceleration_scale": 9.81863788258315,
"normalized_peak_jerk": 18.188645899504852,
"dynamic_acceleration_rms_relative": 0.07444476304415548,
"normalized_jerk_rms": 1.1421572514908371,
"spectral_arc_length": -2.458480081869371,
"spectral_arc_length_rough... | You are an expert six-axis IMU evidence analyst. You receive
twelve panels made from one 12-second accelerometer-and-gyroscope window by an
IMU-JEPA encoder, plus an optional anonymous temporal reference marker.
Produce a rigorous, panel-grounded explanation of observable motion. Integrate
raw morphology, jerk, burst ... | {
"reasoning_chain": {
"step1_signal_quality": "Panel 1 shows continuous, finite values across all six axes. No clipping, saturation, or dropouts are evident, though some minor variation is seen before the main event.",
"step2_acceleration_and_jerk": "Panel 2 displays a long period of near-flatness followed b... | ## Evidence reasoning
1. Panel 1 shows continuous, finite values across all six axes. No clipping, saturation, or dropouts are evident, though some minor variation is seen before the main event.
2. Panel 2 displays a long period of near-flatness followed by a sharp spike near 10 seconds. The raw peak jerk is 203.9 devi... | imu_jepa_panel_insights_v11 | brief_bursts | moderate | single_impulse | ballistic_single_adjustment | 0.85 | standard | true | {
"model": "gemini-3.1-pro-preview",
"passes": 1,
"thinking_level": "medium",
"max_output_tokens": 16000,
"self_consistency_checked": false,
"categorical_agreement": null,
"is_human_ground_truth": false,
"provenance": "gemini_panel_grounded_teacher_hypothesis",
"schema_version": "imu_jepa_panel_insigh... | |
hapt_representation_only:user01:exp01_user01:waist_phone:0 | hapt_representation_only | user01 | exp01_user01 | hapt_smartphone | non_game_motion_source | waist_phone | 50 | false | no_event_marker | null | [
"p1_imu_attn",
"p2_accel_jerk",
"p3_burst_decomp",
"p4_energy_env",
"p5_jerk_timeline",
"p6_gyro",
"p7_spectrum",
"p8_spectrogram",
"p9_importance",
"p10_head_dispersion",
"p11_subject_reference_distance",
"p12_motion_tertile_similarity"
] | {
"n_bursts": 9,
"peak_jerk": 23.421412768782087,
"acceleration_scale": 1.031742553416636,
"normalized_peak_jerk": 6.546050929466676,
"dynamic_acceleration_rms_relative": 0.11899252801335479,
"normalized_jerk_rms": 0.8986363545638449,
"spectral_arc_length": -2.854914712618453,
"spectral_arc_length_rough... | You are an expert six-axis IMU evidence analyst. You receive
twelve panels made from one 12-second accelerometer-and-gyroscope window by an
IMU-JEPA encoder, plus an optional anonymous temporal reference marker.
Produce a rigorous, panel-grounded explanation of observable motion. Integrate
raw morphology, jerk, burst ... | {
"reasoning_chain": {
"step1_signal_quality": "Panel 1 shows continuous, finite traces across all six axes without visible clipping, dropouts, or saturation. The signal quality appears sufficient for analysis without obvious sensor artifacts.",
"step2_acceleration_and_jerk": "Panel 2 displays a clear period ... | ## Evidence reasoning
1. Panel 1 shows continuous, finite traces across all six axes without visible clipping, dropouts, or saturation. The signal quality appears sufficient for analysis without obvious sensor artifacts.
2. Panel 2 displays a clear period of increased acceleration and jerk in the first 5 seconds, follo... | imu_jepa_panel_insights_v11 | brief_bursts | moderate | transition | postural_or_device_repositioning | 0.9 | ambiguous_example | false | {
"model": "gemini-3.1-pro-preview",
"passes": 1,
"thinking_level": "medium",
"max_output_tokens": 16000,
"self_consistency_checked": false,
"categorical_agreement": null,
"is_human_ground_truth": false,
"provenance": "gemini_panel_grounded_teacher_hypothesis",
"schema_version": "imu_jepa_panel_insigh... | |
hapt_marker_motion:user01:exp01_user01:waist_phone:107360 | hapt_marker_motion | user01 | exp01_user01 | hapt_smartphone | non_game_motion_source | waist_phone | 50 | true | video_annotated_postural_transition_onset | 6 | [
"p1_imu_attn",
"p2_accel_jerk",
"p3_burst_decomp",
"p4_energy_env",
"p5_jerk_timeline",
"p6_gyro",
"p7_spectrum",
"p8_spectrogram",
"p9_importance",
"p10_head_dispersion",
"p11_subject_reference_distance",
"p12_motion_tertile_similarity"
] | {
"n_bursts": 9,
"peak_jerk": 6.084585645478407,
"acceleration_scale": 1.0140149583292624,
"normalized_peak_jerk": 4.597525240869555,
"dynamic_acceleration_rms_relative": 0.31371486405533366,
"normalized_jerk_rms": 0.7329465697188561,
"spectral_arc_length": -2.5182617625125383,
"spectral_arc_length_roug... | You are an expert six-axis IMU evidence analyst. You receive
twelve panels made from one 12-second accelerometer-and-gyroscope window by an
IMU-JEPA encoder, plus an optional anonymous temporal reference marker.
Produce a rigorous, panel-grounded explanation of observable motion. Integrate
raw morphology, jerk, burst ... | {
"reasoning_chain": {
"step1_signal_quality": "P1 shows a finite 12-second 6-axis IMU trace with no visible clipping, dropouts, or saturation. The signal appears clean and suitable for analysis.",
"step2_acceleration_and_jerk": "P1 and P2 indicate a period of relative stillness followed by a prominent motion... | ## Evidence reasoning
1. P1 shows a finite 12-second 6-axis IMU trace with no visible clipping, dropouts, or saturation. The signal appears clean and suitable for analysis.
2. P1 and P2 indicate a period of relative stillness followed by a prominent motion event. The raw peak jerk is 6.085 device-units/s and normalized... | imu_jepa_panel_insights_v11 | variable_repeated_movement | moderate | clustered_bursts | intermittent_corrective_control | 0.9 | ambiguous_example | false | {
"model": "gemini-3.1-pro-preview",
"passes": 1,
"thinking_level": "medium",
"max_output_tokens": 16000,
"self_consistency_checked": false,
"categorical_agreement": null,
"is_human_ground_truth": false,
"provenance": "gemini_panel_grounded_teacher_hypothesis",
"schema_version": "imu_jepa_panel_insigh... | |
realworld_representation_only:s10:s10_climbingdown:realworld_chest:0 | realworld_representation_only | s10 | s10_climbingdown | realworld_wearable | non_game_motion_source | realworld_chest | 50.000048 | false | no_event_marker | null | [
"p1_imu_attn",
"p2_accel_jerk",
"p3_burst_decomp",
"p4_energy_env",
"p5_jerk_timeline",
"p6_gyro",
"p7_spectrum",
"p8_spectrogram",
"p9_importance",
"p10_head_dispersion",
"p11_subject_reference_distance",
"p12_motion_tertile_similarity"
] | {
"n_bursts": 13,
"peak_jerk": 308.49340058957637,
"acceleration_scale": 9.696878850322797,
"normalized_peak_jerk": 12.587941438471411,
"dynamic_acceleration_rms_relative": 0.2057141115396967,
"normalized_jerk_rms": 2.5509992922066855,
"spectral_arc_length": -4.633224342984768,
"spectral_arc_length_roug... | You are an expert six-axis IMU evidence analyst. You receive
twelve panels made from one 12-second accelerometer-and-gyroscope window by an
IMU-JEPA encoder, plus an optional anonymous temporal reference marker.
Produce a rigorous, panel-grounded explanation of observable motion. Integrate
raw morphology, jerk, burst ... | {
"reasoning_chain": {
"step1_signal_quality": "Panel 1 displays a continuous, finite trace without obvious clipping, dropout, or saturation. The signal quality appears sufficient for detailed kinematic analysis.",
"step2_acceleration_and_jerk": "Panel 2 shows an initial period of minimal acceleration variati... | ## Evidence reasoning
1. Panel 1 displays a continuous, finite trace without obvious clipping, dropout, or saturation. The signal quality appears sufficient for detailed kinematic analysis.
2. Panel 2 shows an initial period of minimal acceleration variation, transitioning sharply around 5.5s into repeated, moderate-am... | imu_jepa_panel_insights_v11 | variable_repeated_movement | moderate | transition | sustained_repetitive_control | 0.85 | standard | true | {
"model": "gemini-3.1-pro-preview",
"passes": 1,
"thinking_level": "medium",
"max_output_tokens": 16000,
"self_consistency_checked": false,
"categorical_agreement": null,
"is_human_ground_truth": false,
"provenance": "gemini_panel_grounded_teacher_hypothesis",
"schema_version": "imu_jepa_panel_insigh... | |
kuhar_representation_only:p001:1001_A_1:waist_phone:0 | kuhar_representation_only | p001 | 1001_A_1 | kuhar_smartphone | non_game_motion_source | waist_phone | 100 | false | no_event_marker | null | [
"p1_imu_attn",
"p2_accel_jerk",
"p3_burst_decomp",
"p4_energy_env",
"p5_jerk_timeline",
"p6_gyro",
"p7_spectrum",
"p8_spectrogram",
"p9_importance",
"p10_head_dispersion",
"p11_subject_reference_distance",
"p12_motion_tertile_similarity"
] | {
"n_bursts": 14,
"peak_jerk": 3.8501655378283846,
"acceleration_scale": 0.06933347360272277,
"normalized_peak_jerk": 29.99208247906343,
"dynamic_acceleration_rms_relative": 0.6515213483449283,
"normalized_jerk_rms": 7.981133044905269,
"spectral_arc_length": -8.954354677831237,
"spectral_arc_length_roug... | You are an expert six-axis IMU evidence analyst. You receive
twelve panels made from one 12-second accelerometer-and-gyroscope window by an
IMU-JEPA encoder, plus an optional anonymous temporal reference marker.
Produce a rigorous, panel-grounded explanation of observable motion. Integrate
raw morphology, jerk, burst ... | {
"reasoning_chain": {
"step1_signal_quality": "Panel 1 displays a continuous 12-second six-axis trace with no visible clipping, dropout, or non-finite values. The signal morphology indicates clean, uninterrupted data collection without severe artifacts.",
"step2_acceleration_and_jerk": "Panels 1 and 2 reveal... | ## Evidence reasoning
1. Panel 1 displays a continuous 12-second six-axis trace with no visible clipping, dropout, or non-finite values. The signal morphology indicates clean, uninterrupted data collection without severe artifacts.
2. Panels 1 and 2 reveal continuous, low-amplitude oscillatory acceleration across the f... | imu_jepa_panel_insights_v11 | sustained_movement | low | rhythmic_repeated | sustained_repetitive_control | 0.9 | standard | true | {
"model": "gemini-3.1-pro-preview",
"passes": 1,
"thinking_level": "medium",
"max_output_tokens": 16000,
"self_consistency_checked": false,
"categorical_agreement": null,
"is_human_ground_truth": false,
"provenance": "gemini_panel_grounded_teacher_hypothesis",
"schema_version": "imu_jepa_panel_insigh... | |
opportunity_representation_only:S1:S1-ADL1:opportunity_back:0 | opportunity_representation_only | S1 | S1-ADL1 | opportunity_body_imu | non_game_motion_source | opportunity_back | 30 | false | no_event_marker | null | ["p1_imu_attn","p2_accel_jerk","p3_burst_decomp","p4_energy_env","p5_jerk_timeline","p6_gyro","p7_sp(...TRUNCATED) | {"n_bursts":5,"peak_jerk":19262.52670119718,"acceleration_scale":1003.5080440410734,"normalized_peak(...TRUNCATED) | "You are an expert six-axis IMU evidence analyst. You receive\ntwelve panels made from one 12-second(...TRUNCATED) | {"reasoning_chain":{"step1_signal_quality":"Panel 1 displays complete, finite traces across all six (...TRUNCATED) | "## Evidence reasoning\n1. Panel 1 displays complete, finite traces across all six axes for the enti(...TRUNCATED) | imu_jepa_panel_insights_v11 | variable_repeated_movement | high | clustered_bursts | intermittent_corrective_control | 0.85 | high_value | true | {"model":"gemini-3.1-pro-preview","passes":1,"thinking_level":"medium","max_output_tokens":16000,"se(...TRUNCATED) | [{"src":"https://huggingface.co/proxy/datasets-server.huggingface.co/assets/immersely/imu-motion-vlm/--/{dataset_git_revi(...TRUNCATED) |
biraffe2_event_motion:SUB103:SUB103-Gamepad:gamepad:1299240 | biraffe2_event_motion | SUB103 | SUB103-Gamepad | biraffe2_gamepad | biraffe2_lab_game | gamepad | 50 | true | reward | 6.002657 | ["p1_imu_attn","p2_accel_jerk","p3_burst_decomp","p4_energy_env","p5_jerk_timeline","p6_gyro","p7_sp(...TRUNCATED) | {"n_bursts":18,"peak_jerk":39.54225898606825,"acceleration_scale":1.0726171715973853,"normalized_pea(...TRUNCATED) | "You are an expert six-axis IMU evidence analyst. You receive\ntwelve panels made from one 12-second(...TRUNCATED) | {"reasoning_chain":{"step1_signal_quality":"Panel 1 displays a complete, uninterrupted 12-second tra(...TRUNCATED) | "## Evidence reasoning\n1. Panel 1 displays a complete, uninterrupted 12-second trace across all six(...TRUNCATED) | imu_jepa_panel_insights_v11 | sustained_movement | high | sustained | sustained_repetitive_control | 0.85 | ambiguous_example | false | {"model":"gemini-3.1-pro-preview","passes":1,"thinking_level":"medium","max_output_tokens":16000,"se(...TRUNCATED) | [{"src":"https://huggingface.co/proxy/datasets-server.huggingface.co/assets/immersely/imu-motion-vlm/--/{dataset_git_revi(...TRUNCATED) |
hapt_representation_only:user01:exp01_user01:waist_phone:1500 | hapt_representation_only | user01 | exp01_user01 | hapt_smartphone | non_game_motion_source | waist_phone | 50 | false | no_event_marker | null | ["p1_imu_attn","p2_accel_jerk","p3_burst_decomp","p4_energy_env","p5_jerk_timeline","p6_gyro","p7_sp(...TRUNCATED) | {"n_bursts":7,"peak_jerk":23.45706998189944,"acceleration_scale":1.0319594193864217,"normalized_peak(...TRUNCATED) | "You are an expert six-axis IMU evidence analyst. You receive\ntwelve panels made from one 12-second(...TRUNCATED) | {"reasoning_chain":{"step1_signal_quality":"Panel 1 shows continuous, finite values across all six a(...TRUNCATED) | "## Evidence reasoning\n1. Panel 1 shows continuous, finite values across all six axes without visib(...TRUNCATED) | imu_jepa_panel_insights_v11 | brief_bursts | moderate | clustered_bursts | ballistic_single_adjustment | 0.9 | ambiguous_example | false | {"model":"gemini-3.1-pro-preview","passes":1,"thinking_level":"medium","max_output_tokens":16000,"se(...TRUNCATED) | [{"src":"https://huggingface.co/proxy/datasets-server.huggingface.co/assets/immersely/imu-motion-vlm/--/{dataset_git_revi(...TRUNCATED) |
mhealth_representation_only:mHealth_subject10:mHealth_subject10:mhealth_lower_arm:0 | mhealth_representation_only | mHealth_subject10 | mHealth_subject10 | mhealth_imu | non_game_representation_source | mhealth_lower_arm | 50 | false | no_event_marker | null | ["p1_imu_attn","p2_accel_jerk","p3_burst_decomp","p4_energy_env","p5_jerk_timeline","p6_gyro","p7_sp(...TRUNCATED) | {"n_bursts":11,"peak_jerk":197.48412239390746,"acceleration_scale":9.691044516199419,"normalized_pea(...TRUNCATED) | "You are an expert six-axis IMU evidence analyst. You receive\ntwelve panels made from one 12-second(...TRUNCATED) | {"reasoning_chain":{"step1_signal_quality":"P1 shows finite traces for all six axes without visible (...TRUNCATED) | "## Evidence reasoning\n1. P1 shows finite traces for all six axes without visible clipping, dropout(...TRUNCATED) | imu_jepa_panel_insights_v11 | variable_repeated_movement | high | transition | sustained_repetitive_control | 0.9 | standard | true | {"model":"gemini-3.1-pro-preview","passes":1,"thinking_level":"medium","max_output_tokens":16000,"se(...TRUNCATED) | [{"src":"https://huggingface.co/proxy/datasets-server.huggingface.co/assets/immersely/imu-motion-vlm/--/{dataset_git_revi(...TRUNCATED) |
hapt_marker_motion:user01:exp01_user01:waist_phone:129740 | hapt_marker_motion | user01 | exp01_user01 | hapt_smartphone | non_game_motion_source | waist_phone | 50 | true | video_annotated_postural_transition_onset | 6 | ["p1_imu_attn","p2_accel_jerk","p3_burst_decomp","p4_energy_env","p5_jerk_timeline","p6_gyro","p7_sp(...TRUNCATED) | {"n_bursts":6,"peak_jerk":14.2851273916965,"acceleration_scale":0.9960387062756737,"normalized_peak_(...TRUNCATED) | "You are an expert six-axis IMU evidence analyst. You receive\ntwelve panels made from one 12-second(...TRUNCATED) | {"reasoning_chain":{"step1_signal_quality":"Panel 1 displays a complete, continuous six-axis IMU tra(...TRUNCATED) | "## Evidence reasoning\n1. Panel 1 displays a complete, continuous six-axis IMU trace with no visibl(...TRUNCATED) | imu_jepa_panel_insights_v11 | sustained_movement | high | transition | sustained_repetitive_control | 0.9 | ambiguous_example | false | {"model":"gemini-3.1-pro-preview","passes":1,"thinking_level":"medium","max_output_tokens":16000,"se(...TRUNCATED) | [{"src":"https://huggingface.co/proxy/datasets-server.huggingface.co/assets/immersely/imu-motion-vlm/--/{dataset_git_revi(...TRUNCATED) |
End of preview. Expand in Data Studio
README.md exists but content is empty.
- Downloads last month
- 13