AR Glass App for Real-Time Video Processing
We’ve developed a user interface designed for AR glasses connected to the client AI-powered biometric vision model which recognizes people and annotates them on live camera streaming.
01
About
A multinational corporation specializing in advanced IT, networking, and communication solutions for enterprises and governments, with a focus on innovation and cutting-edge technologies.
02
The challenge
The key challenge was to develop an Android app for Vuzix M400 AR glasses that seamlessly connected a user interface designed for AR glasses with an AI-powered biometric vision model. This model, running on the client’s GPU unit within the same Wi-Fi network, needed to provide real-time updates directly displayed on the glasses. It was designed to recognize people from a client database in real time. As a team we’ve never got access to real ML model and real data, so we needed to mock our own server based on client specifications.
03
Goals
Enable real-time video streaming from the AR glasses to a client app via H264/WebRTC.
Display results from the biometric model, including bounding boxes and recognition data, on the AR glasses.
Allow interactive actions on notifications directly through the AR glasses interface.
Adapt the solution for other AR glasses, such as Digilens, and improve performance and configurability.
Low latency, high performance.
04
What we did
We created an optimized Android app for Vuzix M400 AR glasses, enabling efficient GPU communication, seamless data handling, and enhanced performance. The app was tailored for compatibility with other AR devices and focused on improving local processing.
Built a robust Android (Unity) app for Vuzix M400 glasses to interface with the GPU unit over Wi-Fi, enabling seamless communication.
Implemented JSON-based data handling for bounding box rendering and recognition results.
Conducted extensive optimizations for performance, video quality, and processing speed.
Adjusted settings and UI for compatibility with other AR glasses models.
Removed the need for reverse video streaming to focus on local processing.
05
Results
The project delivered seamless integration of AR technology with AI, enabling real-time updates, enhanced user experience, and multi-platform compatibility, all achieved within an impressive 15-day timeline.
Successfully integrated the AR glasses UI with the AI biometric vision model, achieving real-time updates with minimal latency.
Improved user experience with customizable settings and optimized performance.
Delivered a flexible solution compatible with second AR glasses platform.
Identification of as many as 40 individuals in a single video frame simultaneously.
Delivery of ultra-fast performance with a latency of only 50 milliseconds.
The first Proof of Concept was completed and delivered in just 15 days.
06
Profit for the client
The first Proof of Concept was delivered successfully in an exceptionally short period of just three weeks.
The client was able to successfully demonstrate the solution at international fairs.
The optimized system increased operational efficiency and improved the client’s offering in the AR technology space, strengthening their market position.
Ready
to ship faster?
