This Altera-Baidu demo (booth #215B) demonstrates how much faster image classification can be performed using FPGA-accelerated CNNs. For key search functions such as image classification and recognition tasks, CNNs are considered the most advanced technology, capable of achieving previously impossible levels of accuracy. By utilizing Altera Stratix® V FPGAs and the Altera SDK for OpenCL™ (which achieved Khronos OpenCL compatibility certification in May 2013), Baidu has been able to significantly simplify the implementation of parallel processing applications.

Dr. Ren Wu, a distinguished researcher at Baidu, stated, “Baidu is leading the field as a pioneer and leader in deep learning and heterogeneous computing, and FPGA acceleration holds incredible potential. OpenCL support is a game changer that signals a paradigm shift in related fields and will enable FPGAs to expand their reach into mainstream heterogeneous computing.” "By doing so, we are opening the door to infinite possibilities," he said.

Altera's data sensor solutions are based on high-performance Stratix® V and Arria® 10 FPGAs, as well as the next-generation Stratix 10 FPGAs and SoCs. The Stratix 10 FPGA and SoC products are manufactured using Intel's 14nm Tri-Gate process and feature Altera's high-performance HyperFlex™ architecture. Altera's FPGAs provide the high performance and flexibility required for demanding data center environments by combining on-chip memory and DSP blocks with unprecedented reconfigurable logic.

Michael Strickland, Director of Computation and Storage at Altera, said, “Baidu and Altera have introduced a powerful heterogeneous computing solution for accelerating CNN algorithms. Altera FPGAs provide excellent configurability and capabilities such as hard IEEE 754 floating-point multipliers and adders, enabling servers and data centers to quickly meet the ever-changing and complex demands of search, big data, and deep learning and to evolve with the flow of change.”