LunaDream74 commited on
Commit
6060a5e
·
verified ·
1 Parent(s): cd1668e

docs: dataset card with nuScenes attribution

Browse files
Files changed (1) hide show
  1. README.md +16 -5
README.md CHANGED
@@ -41,11 +41,22 @@ thumbs.tar -> repo root <- fetch this one
41
  | `*.jpg` | 70 MB | the same thumbnails loose, 320×180, downscaled from 1600×900 |
42
 
43
  **The thumbnails are redacted.** Faces and licence plates are mosaicked at source resolution before
44
- the downscale, from the person and vehicle boxes of a Faster R-CNN pass — 30,114 regions across the
45
- corpus, 82.8% of frames carrying at least one. Coverage is that detector's recall at a 0.5
46
- confidence floor, so this is a mitigation and not a guarantee that no face is visible; nuScenes'
47
- own blurring pass sits underneath it. `scripts/anonymise.py` in the source repo has the geometry and
48
- the full list of limits.
 
 
 
 
 
 
 
 
 
 
 
49
 
50
  **Fetch `thumbs.tar`, not the loose files.** They hold identical bytes, and the loose copies exist so
51
  the dataset is browsable here — but pulling 6,858 files one at a time is what the Hub rate-limits an
 
41
  | `*.jpg` | 70 MB | the same thumbnails loose, 320×180, downscaled from 1600×900 |
42
 
43
  **The thumbnails are redacted.** Faces and licence plates are mosaicked at source resolution before
44
+ the downscale — 5,511 regions across the corpus, 2,616 frames (38.1%) carrying at least one.
45
+
46
+ Regions come from two sources unioned together: dedicated face and licence-plate detectors
47
+ (YOLO11n, 0.35 confidence floor) contributing 239 faces and 348 plates, and the top 28% of every
48
+ `person` box contributing 4,924 head bands. Vehicle boxes produce no regions. The head band is not a
49
+ fallback for a weak detector — measured on crowd footage the face detector is sound, and 68% of
50
+ frames containing a person still yield no detectable face at all, so the union is what coverage
51
+ actually rests on. Here it is 89% of the redacted area, because nuScenes ships pre-blurred and the
52
+ dedicated detectors fire on only 7.5% of frames.
53
+
54
+ Coverage is therefore the union of two recalls, and this is a mitigation rather than a guarantee
55
+ that no face is visible; nuScenes' own blurring pass sits underneath it. Every frame in this dataset
56
+ is covered by a provenance record naming the exact weights that redacted it — `privacy_detections.jsonl`
57
+ in the source repo's `data/artifacts/`, which `scripts/verify_privacy_assets.py` checks in both
58
+ directions before anything is uploaded. `scripts/anonymise.py` has the geometry and the full list of
59
+ limits.
60
 
61
  **Fetch `thumbs.tar`, not the loose files.** They hold identical bytes, and the loose copies exist so
62
  the dataset is browsable here — but pulling 6,858 files one at a time is what the Hub rate-limits an