Innodisk has released a new AI accelerator card designed for edge devices. The card is built with a third-generation Movidius Myriad X Vision Processing Unit (VPU) from Intel, which is particularly well-suited to computer vision applications like face and object recognition.
According to Innodisk, the AI accelerator card essentially functions like the video card on a standard PC, which handles some of the graphical responsibilities for the computer and minimizes the strain on the CPU. The company notes that many device manufacturers currently opt for low-power CPUs that may not be well-suited to data intensive AI applications.
The Intel VPU is designed for deep neural network inference. The AI accelerator enhances those capabilities, and can generate as much as a 30-fold increase while still minimizing the device’s thermal footprint and power consumption.
Innodisk is releasing two versions of the AI accelerator. The mPCIe features a single Myriad X VPU, while the M.2 2280 features a dual VPU setup. The card is compatible with Windows and Linux systems in addition to the Intel OpenVINO toolkit, which extends the utility of the card to deep learning frameworks like TensorFlow, Apache MXNet, Caffe, and the Open Neural Network Exchange.
Of course, Innodisk isn’t the only company working to develop better computer vision tech for edge devices. Chooch recently released an Edge AI platform that can categorize events on a local video stream, while Visionati unveiled an image recognition API geared towards an enterprise clientele.