HandDiffuse 3aaa125c73ea3ecc batch video-8936933820905412a1e2

episodes/20221208_sequence8/cameras/primary_48/48.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7e69ed9d4815b1575c7b1d6f032c5dc4516ee236cdb0f50f775de3c07f2ee926
+size 162785401

episodes/20221208_sequence8/cameras/primary_48/48.mp4.privacy.json

@@ -0,0 +1,126 @@
+{
+  "schema_version": 1,
+  "video_file": "48.mp4",
+  "source_sha256": "9c8d1cd31d5d1021e7f84d9592d4870f5e314a7a42f33bb03ab5fedd5aece35c",
+  "source_bytes": 317456211,
+  "output_sha256": "7e69ed9d4815b1575c7b1d6f032c5dc4516ee236cdb0f50f775de3c07f2ee926",
+  "output_bytes": 162785401,
+  "processing": {
+    "detectors": [
+      "YuNet_2023mar_.65",
+      "CenterFace_.4",
+      "sideways_90deg_supplement_.65_stride30"
+    ],
+    "detection_stride": 5,
+    "interpolation": "linear bounding-box locations and sizes; every output frame masked",
+    "point_fields": [
+      "time_seconds",
+      "x",
+      "y",
+      "width",
+      "height"
+    ],
+    "boundary_policy": "All dual-detector tracks retained with sparse interpolation and 0.5s boundary extension; residual correction masks are strictly additive. No per-video human review claimed.",
+    "tracks": [
+      {
+        "width": 176,
+        "height": 208,
+        "start": 44.37821154487821,
+        "end": 45.37821154487821,
+        "points": [
+          [
+            44.37821154487821,
+            2175.7322692871094,
+            1477.6214599609375,
+            176,
+            208
+          ],
+          [
+            44.87821154487821,
+            2175.7322692871094,
+            1477.6214599609375,
+            176,
+            208
+          ],
+          [
+            45.37821154487821,
+            2175.7322692871094,
+            1477.6214599609375,
+            176,
+            208
+          ]
+        ],
+        "detections": 1
+      }
+    ],
+    "face_track_count": 1,
+    "codec": "h264",
+    "encoder": "h264_nvenc",
+    "cq": 23,
+    "pixelation_grid": [
+      8,
+      8
+    ],
+    "algorithm": "mano20-alpha-sparse5-v1",
+    "residual_correction_policy": "Additional masks only; original detector-track masks retained.",
+    "hand_protection": {
+      "outward_padding_pixels": 20,
+      "region": "separate convex hulls of recorded projected MANO vertices",
+      "compositing": "opaque original-source alpha inside protection, face mosaic outside",
+      "model_or_camera_refit": false,
+      "missing_motion_frame_indices": [
+        1859
+      ],
+      "missing_pose_policy": "preserve original pixels for entire frame"
+    },
+    "projection": {
+      "episode": "20221208_sequence8",
+      "camera_uid": "primary_48",
+      "parameter_namespace": "primary_original",
+      "original_parameter_id": "48",
+      "calibration_sha256": "815a6e0b416b4d1518ab3c2ec820e6835bbba1e679f594b7e3e495cb2ee5086c",
+      "source_frame_map_sha256": "eb7dc2971e165200d9e714d82df69cd9c0ac073fc0dc363f3299a827360c68d4",
+      "models": {
+        "left": {
+          "motion_sha256": "d0762edc441d324e31b128a1b483ab3fce8406ada84a14a53794b2823ecacb41",
+          "model_sha256": "c4022f7083f2ca7c78b2b3d595abbab52debd32b09d372b16923a801f0ea6a30"
+        },
+        "right": {
+          "motion_sha256": "2b05752d5e14551669e73e74c1332b16d45e4bbe3b4a8ed80fe2ca343a311f91",
+          "model_sha256": "45d60aa3b27ef9107a7afd4e00808f307fd91111e1cfa35afd5c4a62de264767"
+        }
+      },
+      "new_camera_fit": false,
+      "convention": "V_source=V_raw-J_raw[0]+saved_root; X_camera=R*X_source+t; p=K*X_camera",
+      "approximate_mapping_policy": "union of recorded selected and alternative hand projections",
+      "projection_scope": "Consume existing mapping, no new frame-map integrity audit."
+    }
+  },
+  "validation": {
+    "status": "PASSED_HAND_PRIORITY_POLICY",
+    "frames": 1860,
+    "timestamp_equality": true,
+    "dimensions_preserved": true,
+    "audio_streams": 0,
+    "residual_check_frames": 372,
+    "elapsed_seconds": 30.953630094008986,
+    "scope": "Sparse dual-detector residual check, not a proof against all detector blind spots.",
+    "allowed_hand_protection_residuals": [
+      0
+    ]
+  },
+  "visual_review": {
+    "mosaic_visible": true,
+    "hands_reviewed": false,
+    "false_positive_residual_count": 0,
+    "scope": "Automatic exact frame/PTS/dimension/audio checks and offset sparse dual-face detection; approved MANO20 alpha compositor. No per-video human visual review claimed.",
+    "kind": "automated_policy_checks"
+  },
+  "dimensions": [
+    3840,
+    2160
+  ],
+  "frames": 1860,
+  "source_time_base": "1/11988",
+  "timestamps_sha256": "045c86a0bc78c89f4bccf138b0d1fe3e4060396a201b39a170fa89f660e97472"
+}