OPI x Xbox: First Fingernail Segmentation AR Experience
Challenge
OPI sought to connect with gamers through their Xbox collaboration, but existing AR beauty technology couldn't deliver the precision needed for fingernail experiences. Computer vision models existed for faces and hand gestures, but fingernail segmentation presented unique challenges—the surface area was extremely small, requiring millimeter-level precision to look convincing.
The technical problem: Train a custom ML model to identify and track fingernails accurately across multiple angles and positions in real-time mobile AR.
Innovation
We developed the first fingernail segmentation ML model for Snapchat using Snap ML Workbooks to train a custom computer vision system.
Technical breakthrough:
- Produced 50,000+ annotated fingernail training images
- Iteratively refined model performance through multiple learning cycles
- Solved perspective bias by expanding side-view training datasets when accuracy varied by angle
- Achieved real-time precision tracking across nail positions
My Role: As Head of Production at Camera IQ, I led project strategy, team coordination, and technical implementation.
Implementation:
- Collaborated with Creative Director on project brief and strategy
- Managed Creative Technologist building Snap ML Workbooks integration
- Coordinated offshore annotation team for massive training dataset
- Navigated technical roadblocks with Snap partnership support
Results & Impact
Performance:
- Successfully launched the first fingernail segmentation AR experience on Snapchat
- Significantly higher usage than typical beauty AR filters
- Broader demographic appeal extending beyond gaming audiences
- Press coverage for both innovative collaboration and breakthrough AR technology
Strategic Impact:
- Deepened Camera IQ relationships with Snap and Microsoft/Xbox
- Established Camera IQ as leader in custom ML model development for AR
- Created proprietary technology that remains largely unmatched in the market due to training complexity