Datasets:

Alaaharoun
/

QualityVision-walking-sample

+# Dataset export
+- **schema_version**: 1.0.0
+- **data.jsonl**: one JSON object per line — `video_source_id` (job id), `original_filename` (upload name), `keypoints` (primary person), optional `persons[]` with `person_id` + `keypoints` for multi-person frames, `person_count`, `timestamp_ms`, `action_label`, optional `layer11`. With **HQ export**, `original_frame_id` links to the pre-filter id; `source_video_fps` is the source clip frame rate.
+- **low_quality_frames.jsonl** (HQ only): frames that failed the filter, with `reject_reasons` and `quality_metrics`.
+- **export_quality_report.json**: thresholds and formula references for this job.
+- **features.json**: sequence-level aggregates; **bulk** jobs use schema 1.3.x with `videos[]` + `aggregate`.
+- **global_stats.json**: job roll-up; **bulk** adds `totals`, per-video rows under `videos`; **HQ** adds `quality_filter`.
+- **manifest.json**: job metadata and augmentation tags.
+- Verify integrity with `SHA256SUMS`.

With Original videos/bulk_3_videos_de077328_dataset/dataset/SHA256SUMS ADDED Viewed

	@@ -0,0 +1,7 @@

+8211e44ab55fd3092f945239e69f64e66724dc3154e93482dbca1015ae595869  data.jsonl
+75e3e73bab88565d0348612c37d09a207b863ddfbb421f601dbc59a3cadb019e  manifest.json
+f84b20c7a284681c903e944920d5e1e1895c22bdb8e354547704b8c25c6920d3  features.json
+5da021d8cf32c523e522feb335e804a92d99b7750778d99eac0623ca52bb116c  global_stats.json
+689544cbff0a2e7e177c62e0093a31a512a6c5c4095391bcfd4e2d80fc645556  README_dataset.md
+d0376d1043735fbc0e4f421565b1d8f5c0584cfee40c62ea63414bb96d9c4f45  export_quality_report.json
+e473aa569e61f985debce704b7806a84536da84e64bf98d5ff8596b878624119  low_quality_frames.jsonl

With Original videos/bulk_3_videos_de077328_dataset/dataset/data.jsonl ADDED Viewed

The diff for this file is too large to render. See raw diff

With Original videos/bulk_3_videos_de077328_dataset/dataset/export_quality_report.json ADDED Viewed

	@@ -0,0 +1,36 @@

+{
+  "job_id": "bulk_3_videos_de077328",
+  "hq_export": true,
+  "thresholds": {
+    "min_frame_quality_score": 0.4,
+    "min_avg_landmark_visibility": 0.58,
+    "min_lower_body_visibility": 0.42,
+    "min_motion_local": 0.12
+  },
+  "filter_report": {
+    "enabled": true,
+    "thresholds": {
+      "min_frame_quality_score": 0.4,
+      "min_avg_landmark_visibility": 0.58,
+      "min_lower_body_visibility": 0.42,
+      "min_motion_local": 0.12
+    },
+    "frames_total": 492,
+    "frames_accepted": 266,
+    "frames_rejected": 226,
+    "accepted_percentage": 54.07,
+    "quality_score_formula": "Per frame: frame_quality_score = 0.5 * avg_landmark_visibility + 0.5 * motion_local. avg_landmark_visibility = mean(visibility) over all keypoints in the frame. motion_local = stability of consecutive hip-center displacements within the same source video (1.0 at clip edges; lower when step-to-step motion is irregular). Sequence-level quality_score in features.json uses the same visibility mean plus motion_consistency from the full hip displacement series (see features_compute).",
+    "mean_metrics_accepted_frames": {
+      "mean_frame_quality_score": 0.6321,
+      "mean_avg_landmark_visibility": 0.7601,
+      "mean_lower_body_visibility": 0.7443,
+      "mean_motion_local": 0.5042
+    },
+    "quality_score_mean_filtered": 0.6321,
+    "mean_avg_landmark_visibility_filtered": 0.7601
+  },
+  "documentation": {
+    "sequence_quality_score": "Sequence quality_score = 0.5 * avg_landmark_visibility + 0.5 * motion_consistency, where motion_consistency derives from hip-center displacement mean vs std across the exported sequence.",
+    "per_frame_scores": "Per frame: frame_quality_score = 0.5 * avg_landmark_visibility + 0.5 * motion_local. avg_landmark_visibility = mean(visibility) over all keypoints in the frame. motion_local = stability of consecutive hip-center displacements within the same source video (1.0 at clip edges; lower when step-to-step motion is irregular). Sequence-level quality_score in features.json uses the same visibility mean plus motion_consistency from the full hip displacement series (see features_compute)."
+  }
+}

With Original videos/bulk_3_videos_de077328_dataset/dataset/features.json ADDED Viewed

	@@ -0,0 +1,83 @@

+{
+  "schema_version": "1.3.1",
+  "job_id": "bulk_3_videos_de077328",
+  "batch": true,
+  "video_count": 3,
+  "videos": [
+    {
+      "original_filename": "3191808-uhd_3840_2160_25fps.mp4",
+      "fps": 25.0,
+      "exported_frames": 91,
+      "sequence": {
+        "frame_count": 91,
+        "action_label": "running",
+        "avg_landmark_visibility": 0.714905,
+        "hip_center_displacement_mean_norm": 0.043485,
+        "hip_center_path_sum_norm": 3.913618,
+        "inter_frame_steps": 90,
+        "velocity_proxy_available": true,
+        "velocity_mean": 0.043485,
+        "velocity_std": 0.029054,
+        "acceleration_mean": 0.026816,
+        "avg_stride_length": 0.118909,
+        "body_height_norm": 0.697328,
+        "arm_swing_amplitude": 0.243035,
+        "step_frequency": 24.725275,
+        "motion_consistency": 0.331847
+      }
+    },
+    {
+      "original_filename": "3191899-uhd_3840_2160_25fps.mp4",
+      "fps": 25.0,
+      "exported_frames": 78,
+      "sequence": {
+        "frame_count": 78,
+        "action_label": "running",
+        "avg_landmark_visibility": 0.846667,
+        "hip_center_displacement_mean_norm": 0.018343,
+        "hip_center_path_sum_norm": 1.412387,
+        "inter_frame_steps": 77,
+        "velocity_proxy_available": true,
+        "velocity_mean": 0.018343,
+        "velocity_std": 0.011044,
+        "acceleration_mean": 0.008817,
+        "avg_stride_length": 0.098317,
+        "body_height_norm": 0.414525,
+        "arm_swing_amplitude": 0.11369,
+        "step_frequency": 24.679487,
+        "motion_consistency": 0.397884
+      }
+    },
+    {
+      "original_filename": "3209698-uhd_3840_2160_25fps.mp4",
+      "fps": 25.0,
+      "exported_frames": 97,
+      "sequence": {
+        "frame_count": 97,
+        "action_label": "running",
+        "avg_landmark_visibility": 0.732835,
+        "hip_center_displacement_mean_norm": 0.042379,
+        "hip_center_path_sum_norm": 4.068405,
+        "inter_frame_steps": 96,
+        "velocity_proxy_available": true,
+        "velocity_mean": 0.042379,
+        "velocity_std": 0.065743,
+        "acceleration_mean": 0.026001,
+        "avg_stride_length": 0.043413,
+        "body_height_norm": 0.577141,
+        "arm_swing_amplitude": 0.330095,
+        "step_frequency": 24.742268,
+        "motion_consistency": 0.0
+      }
+    }
+  ],
+  "aggregate": {
+    "total_frames": 266,
+    "action_label": "running",
+    "stride": 2,
+    "mean_quality_score": 0.4951,
+    "mean_avg_landmark_visibility": 0.76008,
+    "fps_mean": 25.0
+  },
+  "notes": "Batch export: `videos[]` holds one sequence block per source file. Per-video motion metrics are isolated (no cross-video hip chain). data.jsonl concatenates all frames with unique frame_id and source_video_index. Per-frame HQ filter: see export_quality_report.json and global_stats.quality_filter."
+}

With Original videos/bulk_3_videos_de077328_dataset/dataset/global_stats.json ADDED Viewed

	@@ -0,0 +1,77 @@

+{
+  "schema_version": "1.3.1",
+  "job_id": "bulk_3_videos_de077328",
+  "batch": true,
+  "video_count": 3,
+  "totals": {
+    "exported_frames": 266,
+    "quality_score_mean": 0.4951,
+    "mean_avg_landmark_visibility": 0.76008,
+    "fps_mean": 25.0,
+    "stride": 2,
+    "action_distribution": {
+      "running": 266
+    }
+  },
+  "videos": [
+    {
+      "original_filename": "3191808-uhd_3840_2160_25fps.mp4",
+      "exported_frames": 91,
+      "fps": 25.0,
+      "stride": 2,
+      "quality_score": 0.5234,
+      "estimated_source_duration_sec": 3.64,
+      "avg_sequence_length": 91,
+      "action_label": "running"
+    },
+    {
+      "original_filename": "3191899-uhd_3840_2160_25fps.mp4",
+      "exported_frames": 78,
+      "fps": 25.0,
+      "stride": 2,
+      "quality_score": 0.6223,
+      "estimated_source_duration_sec": 3.12,
+      "avg_sequence_length": 78,
+      "action_label": "running"
+    },
+    {
+      "original_filename": "3209698-uhd_3840_2160_25fps.mp4",
+      "exported_frames": 97,
+      "fps": 25.0,
+      "stride": 2,
+      "quality_score": 0.3664,
+      "estimated_source_duration_sec": 3.88,
+      "avg_sequence_length": 97,
+      "action_label": "running"
+    }
+  ],
+  "artifacts": {
+    "per_frame_records": "dataset/data.jsonl",
+    "sequence_features": "dataset/features.json",
+    "job_manifest": "dataset/manifest.json"
+  },
+  "quality_filter": {
+    "enabled": true,
+    "thresholds": {
+      "min_frame_quality_score": 0.4,
+      "min_avg_landmark_visibility": 0.58,
+      "min_lower_body_visibility": 0.42,
+      "min_motion_local": 0.12
+    },
+    "frames_total": 492,
+    "frames_accepted": 266,
+    "frames_rejected": 226,
+    "accepted_percentage": 54.07,
+    "quality_score_formula": "Per frame: frame_quality_score = 0.5 * avg_landmark_visibility + 0.5 * motion_local. avg_landmark_visibility = mean(visibility) over all keypoints in the frame. motion_local = stability of consecutive hip-center displacements within the same source video (1.0 at clip edges; lower when step-to-step motion is irregular). Sequence-level quality_score in features.json uses the same visibility mean plus motion_consistency from the full hip displacement series (see features_compute).",
+    "mean_metrics_accepted_frames": {
+      "mean_frame_quality_score": 0.6321,
+      "mean_avg_landmark_visibility": 0.7601,
+      "mean_lower_body_visibility": 0.7443,
+      "mean_motion_local": 0.5042
+    },
+    "quality_score_mean_filtered": 0.6321,
+    "mean_avg_landmark_visibility_filtered": 0.7601,
+    "sequence_quality_score_formula": "Sequence quality_score = 0.5 * avg_landmark_visibility + 0.5 * motion_consistency, where motion_consistency derives from hip-center displacement mean vs std across the exported sequence.",
+    "per_frame_quality_formula": "Per frame: frame_quality_score = 0.5 * avg_landmark_visibility + 0.5 * motion_local. avg_landmark_visibility = mean(visibility) over all keypoints in the frame. motion_local = stability of consecutive hip-center displacements within the same source video (1.0 at clip edges; lower when step-to-step motion is irregular). Sequence-level quality_score in features.json uses the same visibility mean plus motion_consistency from the full hip displacement series (see features_compute)."
+  }
+}

With Original videos/bulk_3_videos_de077328_dataset/dataset/low_quality_frames.jsonl ADDED Viewed

The diff for this file is too large to render. See raw diff

With Original videos/bulk_3_videos_de077328_dataset/dataset/manifest.json ADDED Viewed

	@@ -0,0 +1,22 @@

+{
+  "schema_version": "1.0.0",
+  "job_id": "bulk_3_videos_de077328",
+  "video_filename": "bulk_3_videos",
+  "fps": 25.0,
+  "frame_count_exported": 266,
+  "stride": 2,
+  "action_label": "running",
+  "augmentations_applied": [],
+  "license": "proprietary_or_user_provided",
+  "consent": true,
+  "notes": "Generated by Quality Vision Motion Dataset Engine (layer11_pose). Per-frame pose keypoints and Layer 1.1 fields: dataset/data.jsonl. Sequence aggregates: dataset/features.json, dataset/global_stats.json. Bulk: features.json / global_stats.json list per-video metrics plus aggregate totals. HQ export: only frames passing visibility/motion thresholds are in data.jsonl; rejected frames are listed in dataset/low_quality_frames.jsonl. See export_quality_report.json for formulas and thresholds.",
+  "dataset_format": "pose_estimation_sequence_dataset",
+  "export_manifest_version": "1.1",
+  "bulk_export": true,
+  "source_video_filenames": [
+    "3191808-uhd_3840_2160_25fps.mp4",
+    "3191899-uhd_3840_2160_25fps.mp4",
+    "3209698-uhd_3840_2160_25fps.mp4"
+  ],
+  "hq_export": true
+}