The face-recognition algorithms use the latest Deep Learning algorithms that provide state-of-the art face-recognition performance. For each face that is recognised, millions of simple computations need to be performed to compute the unique template for that face. To achieve the processing required for face-recognition on live video feeds, these computations need to be performed on relatively new GPU cards. Intel are working hard to make these computations run quicker on CPUs, however they are not efficient enough yet. When they are this requirement might be eased.
Currently, SmartVis Identifier requires a GPU supplied by Nvidia. Nvidia provide efficient libraries for performing Deep Learning. Intel are working hard to write efficient libraries for their Intel graphic chipsets. However, they are not efficient enough yet to be used for real-time face-recognition on HD video streams. Identifier only supports Nvidia GPUs at this time.