Author：Prof. Han Deqiang from Beijing University of Technology
In order to help the participating teams to better select the boards and complete the contest, we have the honor to invite Prof. Han Deqiang from Beijing University of Technology to give some suggestions to the participating teams of the 2018 Intel Cup Embedded System Design Contest. Prof. Han Deqiang, who has many years of experience in embedded systems, is one of the senior judges of the Intel Cup Embedded System Design Contest. We believe that his suggestions will benefit the participating teams.
The following are the suggestions given by Prof. Han Deqiang.
Disclaimer: These suggestions only represent my own point of view and do not represent the views of other judges.
The theme of this year's Intel Cup Embedded System Design Contest is: “Artificial Intelligence and IoT”. The organizing committee provides a set of UP2 IoT development kits (mandatory), a Cyclone 5 FPGA development board (optional), and a Movidius neural compute stick (optional). How to choose from these hardware boards? It is the most important issue before the teams. According to some previous experiences, I will give you some suggestions.
First of all
is the more the boards used, the better the chance of winning? Not really. Unreasonable match will be counterproductive. For an embedded system work, it is necessary to give full play to the maximum performance of the hardware board according to practical applications. The higher the utilization rate, the higher the cost performance, and the higher the application value.
The UP2 board has many powerful functions. In addition to the powerful Apollo Lake processor (N3350), you can also connect various sensor modules via the Arduino-compatible Grove Pi+ expansion board, or you can connect your own developed sensor modules. It supports many programming languages, such as C++, Python, JAVA.
is it necessary to use two additional sets of optional artificial intelligence development kits when AI is involved? Whether hardware acceleration is necessary depends on the specific algorithm used. For applications such as computer vision, multimedia codec, signal and data processing, and machine learning, the UP2 board is fully capable. For example, the UP2 board can use the Intel Math Kernel library to optimize IoT data processing; it can also connect to Microsoft Azure or Amazon Web Services (AWS), transfer data through MQTT, etc.
what can the Cyclone 5 FPGA development board do? In addition to hardware acceleration, Cyclone 5 FPGA development boards also have very rich I/O resources available. Remember to use it reasonable!