Google researchers achieved nearly 91% smile-detection accuracy on the Faces of the World (FotW) benchmark by training racial and gender classifiers as part of their model.
Notes on verification
Confirmed by Google's own research publication page, the original arXiv paper (InclusiveFaceNet), and independent tech press (The Register, TechMonitor) all agreeing on the 91% figure and race/gender classifier methodology. [tier=gold indep_score=0.838 clusters=4 claim_tier=notable]
Sources
- Google’s New AI Smile Detector Shows How Embracing Race and Gender Can Reduce Bias (seed:technology_and_ai)
- https://ai.google/research/pubs/pub46513/ (corroboration)
- https://www.theregister.com/software/2017/12/06/google-learns-to-smile-because-ais-bad-at-it/798222 (corroboration)
- https://www.techmonitor.ai/digital-economy/ai-and-automation/google-ai-diversity (corroboration)
- https://arxiv.org/pdf/1712.00193 (corroboration)