For humans, it goes without saying that vision is extremely valuable. When you stop to think about it, it’s remarkable what a diverse set of capabilities is enabled by human vision – from reading facial expressions, to navigating complex three-dimensional spaces (whether by foot, bicycle, car, or otherwise), to performing intricate tasks like threading a needle.

Processor vendors and system designers share a common concern: how to make sure their products meet customer needs. For processor vendors, a key challenge is to design architectures with enough performance to meet the demands of current and anticipated applications while staying within acceptable power and cost constraints, and enabling good software developer productivity. Fundamentally, processor designers need to bring together the demands of algorithm workloads together with the capabilities of processor architectures.

A variety of visible light- and infrared-based technologies are currently available for determining objects' locations in 3D space, as well as ascertaining their surface shapes and movements. However, these approaches sometimes come with privacy concerns; they also are unable to see inside or through an object. And approaches to seeing inside or through an object, such as conventional and backscatter x-ray systems, along with the millimeter wave scanners now in use in U.S.

Hard on the heels of the public release of CEVA's second-generation convolutional neural network toolset, CDNN2, the company is putting the final touches on its fifth-generation processor core, the CEVA-XM6, designed to run software generated by that toolset.

Based on the pace of investment and acquisitions, and the level of buzz (some would say "hype") surrounding augmented reality and virtual reality, it is obvious that these technologies are hot.

With good reason, I think.

The world of high tech abounds with cautionary tales of decisions based on inadequate information. With 20/20 hindsight, it's easy to see where wrong decisions were made, but when you're in the thick of it, you often fail to see the forest for the trees. Careful technology decision-makers have long relied on BDTI to provide them with the key insights they need to make tough decisions. In addition to benchmarks and in-depth technical analysis, BDTI has been tracking and analyzing business, market, and technology trends in the embedded space for over 15 years.

The general trend of recent times is toward developing DSPs that are specialized for particular "killer" applications, such as wireless or embedded vision. Many other applications also need DSPs, however, and Cadence's Tensilica Fusion DSP core family, which complements the company's more application-specific DSPs, is one example of an architecture that aspires to serve a broader range of general applications.

Last year, when CEVA introduced the initial iteration of its CDNN (CEVA Deep Neural Network) toolset, company officials expressed an aspiration for CDNN to eventually support multiple popular deep learning frameworks. At the time, however, CDNN launched with support only for the well-known Caffe framework, and only for a subset of possible layers and topologies based on it.

If you’ve read recent editions of this column, you know that I believe that embedded vision – enabling devices to understand the world visually – will be a game-changer for many industries.  For humans, vision enables many diverse capabilities: reading your spouse’s facial expression, navigating your car through a parking garage, or threading a needle.  Similarly, embedded vision is now enabling all sorts of devices (from vacuum cleaning robots to cars) to be more autonomous, easier to use, safer, more efficient and more capable.

We're hearing more and more about the effectiveness of deep learning for a growing range of applications. With the surge in the volume of data available for training, and reduction in the cost of computing, technologists have turned to deep neural networks for solutions to compute-intensive applications such as speech and image recognition, 3D object recognition, and natural language processing. Particularly where the virtually unlimited processing power and memory of cloud and enterprise computing can be used, deep learning is extremely effective.