Toshiba Device & Storage executes image recognition using artificial intelligence (AI) deep learning at high speed and low power consumption SoC (System on a Chip) for automotive use. The SoC compared to the previous version is ten times faster in terms of processing speed and four times more efficient.
In image recognition by deep learning, a large number of multiply-add operations must be executed, so it takes time to process with a normal processor, and when processing is performed at high speed, a large amount of power is consumed and to counter this a DNN accelerator was developed that executes image recognition by deep learning on hardware and implementing it on SoC.
The DNN accelerator is equipped with 256 product-sum operation units. By processing computations in parallel, efficiency is improved and thereby improving image recognition.