“We are excited about these results,” said Chris Kachris, CEO of Inaccel, “the limitations faced by CPU and GPU-based facial detection systems will be shattered by applications that utilize FPGA clusters.” The test was conducted on an FPGA cluster operated by VMaccel, a cloud service provider specialising in FPGA-based High-Performance Computing located in Cheyenne, Wyoming.
Object detection in video is a computationally intensive task that requires large amounts of processing power. FPGA-based hardware acceleration is able to provide enhanced processing power to increase throughput and decrease latency. Inaccel offers an integrated framework that allows its clients to utilise the power of FPGA clusters for facial detection and other intensive applications. Together, VMaccel and Inaccel can support deploying these workloads efficiently across hundreds of FPGAs.
Adaptable hardware platforms that can more efficiently offer greater performance and lower latency than GPUs, FPGAs can be used to run any computationally intensive applications such as machine learning, video processing, quantitative finance, or genomics.
However, deployment of applications on FPGAs by users with no prior experience can be challenging. Inaccel provides an FPGA resource manager that allows instant deployment, scaling and resource management. Users can deploy their application using Python, Spark, Jupyter notebooks or even terminals. Inaccel is making its platform available for demonstration purposes to qualified parties.
