Label a sample
Include challenging capture conditions and annotate the sensitive regions independently of the detector.
PII ZERO / EVIDENCE
See the output. Know the limits.
A useful evaluation measures what was missed, what was unnecessarily redacted, and what the workflow costs to run.
PROCESSED DEMONSTRATION
Robotics-lab face redaction.
OriginalPII Zero output
Two researchers. One robot workspace.
DRAG TO COMPARE ↔Public stock footage processed with face detection, tracking and blur. This run does not evaluate text, screens, plates or audio.
| Input | 8 seconds · 1280 × 720 · 25 fps |
|---|---|
| Frames processed | 200 |
| Face tracks | 2 tracks, each present in 200 frames |
| Frames with redaction | 200 / 200 |
| Same-detector residual check | 0 unredacted residual detections reported |
A same-detector re-scan can repeat the original detector’s blind spots. This is not independent ground truth, a recall measurement, or a guarantee of anonymization. Frame coverage measures where a mask was applied, not whether every sensitive region was found.
PILOT PROTOCOL
Measure on your actual distribution.
Measure coverage
Report missed regions, unnecessary masks, localization quality and consistency over time, broken down by PII category.
Measure utility & cost
Review the remaining task signals, end-to-end throughput on target hardware and the amount of human review required.
Screen detection and recapture or liveness classification are separate tasks. The demonstrations here do not establish their accuracy or represent a validated liveness benchmark.
LET'S BUILD YOUR EVALUATION
Start with your footage.
Bring your data requirements. We’ll work through coverage, deployment and an evaluation on a representative sample.
15-minute introduction. No footage upload required.