Welcome to myAshisuto’s blog post series!
What We Do:
We are building smart glasses. We use advanced technologies like hyperbolic 3-manifolds for computationally efficient and ultra-personalized data processing on the edge and between mesh devices. We want to create smart sunglasses that can automatically capture important moments based on biometric data triggers.
We are currently building a demo using data recorded by our partner UP Electromods during a sports bike ride. At myAshisuto, we learn through hands-on experience. In this blog series, we will share our learnings from the process!
As a first step, we are working on overlaying Garmin fitness data onto Insta360 videos. Let us share what we’ve learned from integrating data streams from multiple devices.
Key Lessons:
- Computational Challenges in Data Transformation
Data transfer and processing can be painfully slow when dealing with large video files. For instance, it took us 7 hours to render and overlay a photo sequence onto a 2-hour bike ride video! Streamlining these workflows is essential for efficiency. - File Start Times Are Often Misaligned
That fitness trackers and video cameras rarely start recording at the same time, and synchronization is not straightforward. - Anchor Points Matter
GPS data and time stamps are reliable anchors for synchronising streams. Note that we must be cautious about time zone offsets and device-specific quirks. - Focus on Key Moments, Not Entire Footage
Imagine creating and editing video, not every moment deserves attention. If you can pinpoint a key interest moment, you probably don’t want to sift through every second of the video. Focusing on points of interest creates a better user experience while saving time and resources.
The next step would be to identify the key moments first, and then layer metrics on the part of the video at the identified moments (rather than overlaying metrics on the whole video and trimming key moments later). We will also look into established methods such as time- or GPS-based matching, as synchronization difficulties are known challenges.
We will be sharing our challenges and learnings through our blog in the future. By gathering feedback from potential users and industry experts, we hope to discover new ideas that we can use to improve and refine our solution development. Please share this post with your community and/or let us know what you think!