| Generatable or reverse engineerable personal data? |
Yes, depending on the input. GEM reconstructs body shape and motion from video, and its SOMA outputs can describe real individuals. This concerns information derived from inference inputs; geometric motion output should not be assumed to be anonymous. |
| Personal data used to create this model? |
Partial — the training corpus combines synthetically rendered human-motion video with real-world visual data. Source motion-capture recordings and real-world images or videos can contain information about real individuals. The synthetic nature of rendered training videos does not establish that all underlying source data is non-personal. |
| Was consent obtained for any personal data used? |
Yes — the required consent was obtained for personal data used across the GEM v1.1 training data. |
| Description of methods implemented in data acquisition or processing, if any, to address the prevalence of personal data in the training data: |
Personal data was minimized to what was required for model development. The synthetic portion of the training corpus is rendered from motion-capture data and character and scene assets. The real-world visual subset is automatically annotated for human-motion estimation. Required consent and dataset provenance cover both portions of the GEM v1.1 training data. |
| How often is dataset reviewed? |
Before release. |
| Is a mechanism in place to honor data subject right of access or deletion of personal data? |
Not confirmed for this release. The applicable access or deletion mechanism and the responsible dataset owner or provider require confirmation. |
| If personal data was collected for the development of the model, was it collected directly by NVIDIA? |
Not confirmed by dataset. The training corpus includes NVIDIA in-house motion-capture data and externally released real-world visual data; the original collector of personal data must be identified for each source. |
| If personal data was collected for the development of the model by NVIDIA, do you maintain or have access to disclosures made to data subjects? |
Not confirmed for this release. This depends on which personal data, if any, was collected directly by NVIDIA and the disclosures associated with that collection. |
| If personal data was collected for the development of this AI model, was it minimized to only what was required? |
Yes — personal data was minimized to what was required for model development across the GEM v1.1 training data. |
| Was data from user interactions with the AI model (e.g. user input and prompts) used to train the model? |
No — data from user interactions with the model was not used to train GEM v1.1. |
| Is there provenance for all datasets used in training? |
Yes — provenance is available for all datasets used to train GEM v1.1, including the motion-capture and real-world visual data sources. |
| Does data labeling (annotation, metadata) comply with privacy laws? |
Not confirmed for this release. Annotations are generated automatically, but this does not by itself establish the privacy compliance of the labeling process for all data sources. |
| Is data compliant with data subject requests for data correction or removal, if such a request was made? |
Not confirmed for this release. The handling of any correction or removal requests needs to be confirmed with the responsible dataset owner or provider. |
| Applicable Privacy Policy |
NVIDIA Privacy Policy |
