Smart glasses maker lands 50K palm gesture participants

Overview

A wearable that recognizes the wave of a hand has to learn that wave from real hands in real places.

A Smart Glasses Maker needed a large-scale, synchronized video and IMU gesture dataset ” inclusive across gender, ethnicity, and demographics ” captured at scale in real-world conditions and on aggressive timelines.

Firstsource ran multi-city capture across the USA and India, delivering 50,000+ participants with per-participant signal of 100+ images and 25+ IMU-synced videos across 200+ outdoor and indoor locations.

This was Intelligence that Operates: synchronized multimodal capture in high-footfall public spaces, on a launch cadence.

Challenges

• Lab-only gesture data won't help a wearable that ships into public space. Smart glasses get used at monuments, parks, train stations ” environments lab capture can't replicate. The dataset has to come from where the device will live.
• High-footfall public capture is a permissions and safety problem before it's a data problem. Working with sensitive prototype equipment in busy public locations requires local compliance, safe device handling, and rapid permission cycles.
• Demographic inclusivity at 50,000+ participants is intentional, not opportunistic. Smart glasses gesture models that work for everyone need datasets built for that from the start ” gender, ethnicity, and demographic coverage cannot be sampled after the fact.

How We Made It Happen

We ran multi-city capture as one program with safe device handling, demographic targeting, and same-day quality discipline.

• Multi-city public-space capture across USA and India. 200+ outdoor and indoor locations with strict on-site QC and same-day quality checks.
• Physical AI delivered as a synchronized video + IMU corpus. 50,000+ participants captured with 100+ images and 25+ IMU-synced videos each ” multimodal alignment built into the dataset.
• Inclusivity by design across gender, ethnicity, and demographics. Recruitment matched the user base the model would ship for ” not the population that was easiest to reach.

Conclusion

Smart glasses gesture features only work if their training data was captured where the glasses will be worn. Firstsource ran multi-city public-space capture at scale across two countries ” turning wearable gesture data into Intelligence that Operates.