PlanetSpark's EdgeAI Box based on AMD helps accelerate video analysis

Project Challenges

PlanetSpark believes that the post-epidemic environment will require the deployment of more security cameras for crowd monitoring, traffic monitoring, and building access and security, so it set out to build an AI edge box that can handle large amounts of local video analysis.

Among the must-have features, the box needs to be able to handle multiple video streams simultaneously, and needs to be energy efficient and have good heat dissipation.

Additionally, the solution needed to be customizable.

“Customer needs today vary,” said Joseph Lee, head of venture capital and solutions marketing at PlanetSpark. “That means customization is critical.”

The company decided to address this need with an FPGA-based solution, its first use of FPGAs.

solution

PlanetSpark selected AMD Zynq™ UltraScale+™ MPSoC for its EdgeAI box . FPGA-based solutions enable the high flexibility needed to meet evolving requirements and standards. In addition, they offer ultra-low latency and low power consumption .

“FPGAs are flexible and reprogrammable, so we can customize designs even as AI models continue to evolve,” said David Wang Qing Sheng, vice president of PlanetSpark, a PlanetSpark-backed company. “ AMD’s hybrid solution can adapt to these changing market needs through an FPGA-based architecture combined with multiple processing cores .”

Mr. Wang said that AMD provides both hardware (FPGA architecture) and software (AMD Vitis™ AI development environment) and PlanetSpark now recommends it to developers. “These are extremely important differentiators that help our downstream partners and customers easily start development.”

He added that the Vitis AI model not only helps guide customers in their initial design choices, but also provides an “indispensable library” of APIs and other resources.

Mr. Wang said that the design experience with AMD was very good. He said, "AMD provided strong support from a global team in all aspects of software, hardware and middleware design. AMD has been a great help to our success, not only providing us with good support from a global team during the initial hardware startup process, but also providing us with continuous support when we encounter problems that we have not seen before."

Additionally, Lee noted that AMD’s guidance was crucial to the success of the EdgeAI Box X7. Lee continued, “This was our first time using FPGAs, but the learning curve was short. Fortunately, AMD arranged a lot of global support for our team. This is something we are very grateful for.”

Design Results

PlanetSpark’s EdgeAI boxes help deliver interactive experiences and responses to near-real-time insights, Lee said. The solution is stackable, meaning that in applications such as smart parking, “ users can replace multiple GPUs with EdgeAI boxes and perform video analytics across more cameras with lower latency in a smaller, lower-cost, more power-efficient form factor. ”

“When we think about platforms to align, we think about the longevity of our product lines,” he said. “We want to move forward and not just stay on one product. We need flexibility in our product lines to protect our long-term investments. The knowledge that AMD has imparted to our R&D team is what we want to continue to grow and deepen in FPGAs.”

“AMD has set the gold standard for FPGAs and adaptive computing, and we want to continue to align closely with them,” Lee said.