HandDiffuse 3aaa125c73ea3ecc batch correction-fd52f7c4b4d73e964543-00875093ce19

episodes/20221208_sequence1/cameras/primary_1/1.mp4

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:aaa0ae7a900982668f3b46fb286b320260e452ccb750fcb6e72f5382647c47a6
-size 200255918
+oid sha256:00875093ce19d3a3b50e3b14226875fd4228f7adacb02532ac13a8151fa61a13
+size 200245650

episodes/20221208_sequence1/cameras/primary_1/1.mp4.privacy.json

@@ -3,8 +3,8 @@
   "video_file": "1.mp4",
   "source_sha256": "1a3ad461a9b5ac52853ebd58c89e079da16b03fbaedd12305fcb227f19333d0a",
   "source_bytes": 349883114,
-  "output_sha256": "aaa0ae7a900982668f3b46fb286b320260e452ccb750fcb6e72f5382647c47a6",
-  "output_bytes": 200255918,
+  "output_sha256": "00875093ce19d3a3b50e3b14226875fd4228f7adacb02532ac13a8151fa61a13",
+  "output_bytes": 200245650,
   "processing": {
     "detectors": [
       "YuNet_2023mar_.65",
@@ -19,7 +19,7 @@
       "width",
       "height"
     ],
-    "boundary_policy": "All dual-detector tracks retained with sparse interpolation and 0.5s boundary extension; no per-video human review claimed.",
+    "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": 430,
@@ -2827,13 +2827,6 @@
             270,
             312
           ],
-          [
-            67.6343009676343,
-            1387.590560913086,
-            810.0202484130859,
-            376,
-            268
-          ],
           [
             67.7344010677344,
             1500.8074645996094,
@@ -2870,7 +2863,7 @@
             396
           ]
         ],
-        "detections": 404
+        "detections": 403
       },
       {
         "width": 102,
@@ -2915,9 +2908,39 @@
           ]
         ],
         "detections": 3
+      },
+      {
+        "width": 376,
+        "height": 268,
+        "start": 67.1343009676343,
+        "end": 68.1343009676343,
+        "points": [
+          [
+            67.1343009676343,
+            1387.590560913086,
+            810.0202484130859,
+            376,
+            268
+          ],
+          [
+            67.6343009676343,
+            1387.590560913086,
+            810.0202484130859,
+            376,
+            268
+          ],
+          [
+            68.1343009676343,
+            1387.590560913086,
+            810.0202484130859,
+            376,
+            268
+          ]
+        ],
+        "detections": 1
       }
     ],
-    "face_track_count": 2,
+    "face_track_count": 3,
     "codec": "h264",
     "encoder": "h264_nvenc",
     "cq": 23,
@@ -2926,6 +2949,7 @@
       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",
@@ -2966,7 +2990,7 @@
     "dimensions_preserved": true,
     "audio_streams": 0,
     "residual_check_frames": 409,
-    "elapsed_seconds": 28.694657711996115,
+    "elapsed_seconds": 26.050738100981107,
     "scope": "Sparse dual-detector residual check, not a proof against all detector blind spots.",
     "allowed_hand_protection_residuals": []
   },