The new Innodisk AI accelerator card series provides a highly capable Vision Processing Unit (VPU) for AI inference through machine vision
Taipei – December 19, 2019 – AI is taking the world of IoT by storm. While many aspects of AI are computing and resource-intensive, others rely on small and swift devices that can quickly deliver results. Innodisk’s new AI accelerator cards provide exponential performance gains for AI inference on the edge.
Powered by the Most Advanced Intel® Movidius™ VPU
Innodisk’s new AI accelerator cards use Intel’s third-generation Movidius Myriad™ X Vision Processing Unit (VPU), which is designed for deep neural network inference. As such, the VPU enhances all vision inference applications such as facial recognition, vehicle registration plate recognition, and many other machine vision applications. The results are staggering, enabling an up to a 30-fold boost in performance depending on the edge device’s CPU.
Improving Machine Vision on the Edge
Embedded devices typically found on the edge are ill-equipped for image recognition. The CPU is normally chosen for its low power consumption and thermal efficiency, making it particularly unsuited for AI inference. Much like a video card takes the load off the CPU for graphics rendering in PCs, the VPU assists edge devices in picking out objects and patterns in images and videos. Innodisk’s new AI accelerator cards achieve all this while leaving a small thermal footprint and maintaining low power consumption.
Integrating AI into IoT Applications
Innodisk’s AI accelerator cards have no minimum order quantity and come in two form factors; mPCIe with one Myriad X (EMPA-I101) and M.2 2280 with two Myriad X (EGPA-I201) inside. Windows and Linux are supported, and deep learning frameworks such as Caffe, TensorFlow, Apache MXNet, and Open Neural Network Exchange (ONNX) are also supported through the Intel OpenVINO™ toolkit. The AI accelerator cards are a significant step toward the realization of AIoT and serve as a new cornerstone in Innodisk and its partners’ overall AI architecture.