Graphics IP supplier Imagination Technologies has long advocated the acceleration of edge-based deep learning inference operations via the combination of the company's GPU and ISP cores. Latest-generation graphics architectures from the company continue this trend, enhancing performance and reducing memory bandwidth and capacity requirements in entry-level and mainstream SoCs and systems based on them.

Convolutional neural networks (CNNs) may be the hot artificial intelligence (AI) technology of the moment, in no small part due to their compatibility for both training and inference functions with existing GPUs, FPGAs and DSPs as accelerators, but they're not the only game in town. Witness, for example, Australia-based startup BrainChip Holdings and its alternative proprietary spiking neural network (SNN) technology (Figure 1).

At the Embedded Vision Summit in May, I had the privilege of hearing a brilliant keynote presentation from Professor Jitendra Malik of UC Berkeley.

Benchmarks are an important resource for system and processor designers alike. System designers need good benchmarks to understand how a processor will meet the requirements of their applications and to compare processors accurately. Processor designers need good benchmarks to assess how their processors will perform on their target applications and to prove their processors' capabilities to customers.

Qualcomm is expanding its reference camera module program with three new configurations targeting biometrics and depth-sensing functions in Android-based smartphones, tablets, AR (augmented reality) and VR (virtual reality) headsets, and other devices.

When Movidius unveiled the Fathom Neural Compute Stick, based on its Myriad 2 VPU (vision processor), at the May 2016 Embedded Vision Summit, the company targeted a $99 price tag and initially planned to support the TensorFlow framework, with support for Caffe and other frameworks to follow.

On a recent vacation, I was struck by how indispensable smartphones have become for travelers. GPS-powered maps enable us to navigate unfamiliar cities. Language translation apps help us make sense of unfamiliar languages. Looking for a train, taxi, museum, restaurant, shop or park? A few taps of the screen and you’ve found it.

And yet, there’s a vast amount of useful information that isn’t at our fingertips. Where’s the nearest available parking space? How crowded is that bus, restaurant, or museum right now?

As SoCs become more complex and specialized, incorporating numerous and varied processor cores and dedicated accelerators, it has become more and more difficult to program them. This is particularly true of chips targeting vision-based applications. To meet the performance demands and high data rates of vision applications, vendors are designing heterogeneous devices that incorporate different classes of processors—CPUs, DSPs, GPUs, FPGAs, and special-purpose engines.

AImotive has been developing its aiDrive software suite for advanced driver assistance systems (ADAS) and autonomous vehicles for nearly a decade. As the computing demands of its algorithms continue to increase, the company is finding that conventional processor approaches aren't keeping pace.

Last June, when Synopsys unveiled its latest-generation DesignWare EV6x vision processor core, the company also introduced an 880-MAC, 12-bit convolutional neural network (CNN) companion processor, the CNN880. Although the CNN880 is optional for Synopsys customers using the EV6x, it's been a key factor (often the lead factor, in fact) in greater than 90% of EV6x customer engagements, according to Product Marketing Manager Gordon Cooper.