From Camera to display : HDMI TMDS Video Pipeline design on Pynq z1

Deep neural networks have become a key enabler for intelligent data analysis in applications that demand timely processing and near real-time decision-making. In sensor network environments, outlier detection in time-series data is particularly critical for maintaining data integrity, fault diagnosis, and system reliability. Although DNN-based techniques have demonstrated high detection accuracy, their substantial computational and memory requirements limit their applicability in low-power, low-cost edge devices. These limitations highlight the need for efficient algorithm–hardware co-design strategies that support real-time operation. In this webinar we present a framework to efficiently implement Deep Learning algorithms by exploiting the PYNQ platform from USB host to HDMI video pipeline with accuracy of 98 percent in active resolution of 1920*1080p commonly used in embedded system applications.