Computer Science · Computer Vision and Pattern Recognition
A Glass-to-Glass Testbed: Towards effective Latency Analysis
Anna Semeraro, Carlo Augusto Grazia, Luca Bedogni
See on the mapAbstract
This paper introduces a low-cost and reproducible framework for measuring glass-to-glass (G2G) latency in real-time video systems. Unlike existing solutions, which are often proprietary, expensive, or poorly documented, our approach combines a photodiode, a microcontroller, and lightweight calibration routines to achieve accurate end-to-end latency measurements. The framework is validated across heterogeneous devices, revealing the impact of hardware tiers and video codecs (e.g., H.264 vs. Motion JPEG) on responsiveness. Beyond smartphones, we demonstrate adaptability to complex pipelines such as remote driving and wearable devices, where latency directly affects safety and user experience. Released as an open-source tool, the framework fills a methodological gap in latency research and offers practical guidelines for optimizing multimedia pipelines in domains including virtual reality, telemedicine, and autonomous mobility.