Privacy handling of PhysicalAI-Autonomous-Vehicles: anonymization coverage and official documentation

Hello, We are evaluating the PhysicalAI-Autonomous-Vehicles (driving) dataset internally at our company, under the NVIDIA Autonomous Vehicle Dataset License Agreement. We checked sample frames and confirmed that license plates appear to be blurred.

We also saw a past Hugging Face discussion (#2, by an NVIDIA team member, Oct 2025) saying that the videos were modified to remove PII including faces and license plates, and that `blurred_boxes` are provided. That discussion page is no longer accessible, and we could not find an official statement in the dataset card. Could you please clarify the following?

1. Are faces and license plates anonymized in all clips and all camera views (including the rear and tele cameras)?

2. How was the anonymization done (automatic detection, human review, etc.)? Are its coverage and accuracy documented? Is there a known miss rate?

3. Is there an official document (dataset card, paper, or other) that describes the privacy handling? If not, could one be added?

4. If we find a frame where a face or license plate is still identifiable, how should we handle it? Is there a reporting channel? Thank you.

Hi team, following up on this topic. We have not yet received a response and would appreciate any guidance on the following:

1. Whether faces and license plates are anonymized in all clips and all camera views

2. How the anonymization was done, and any documented coverage or known limitations

3. Whether an official document describing the privacy handling exists (or could be added)

4. The recommended way to report a frame where a face or license plate is still identifiable

If this is not the right place for this question, could you please point us to the appropriate channel? Thank you.