Tag Archives: Computer Vision

Improving Computer Vision with Two AI Processors

Computer vision is becoming a necessity in IoT and automotive applications. Engineers are trying for the next level in computer vision with two AI processors. They hope that two AI processors will help to make computer vision not only more efficient but also more functional.

One of the fastest-growing applications of artificial intelligence, computer vision is jostling for attention between prestigious fields like robotics and autonomous vehicles. In comparison to other artificial intelligence applications, computer vision has to rely more on the underlying hardware, where the underlying imaging systems and processing units overshadow the software performance.

Therefore, engineers are focusing on cutting-edge technology and state-of-the-art developments for the best vision hardware. Two companies, Intuitive and Syntiant, are now making headlines by supporting this move.

Israeli company, Intuitive, recently announced that its NU4000 edge Artificial Intelligence processor will be used by Fukushin Electronics in their new electric cart, POLCAR. The processor will allow the cart to have an integrated obstacle detection unit.

Requiring top performance and power efficiency when operating a sophisticated object detection unit in a battery-powered vehicle like an electric cart made Fukushin use the NU4000. The edge AI processor from Intuitive is a multicore System on a Chip or SoC that can support several on-chip applications. These include computer vision, simultaneous localization and mapping or SLAM, and 3d depth-sensing.

The NU4000 achieves these feats by integrating three Vector Cores that together provide 500 COPS. There is also a dedicated CNN processor offering three CPUs, 2 TOPS, a dedicated SLAM engine, and a dedicated depth processing engine. Intuitive has built this chip with a 12nm process, and it can connect up to two displays and six cameras with an LPDDR4 interface.

With a small form factor and low power consumption, the NU4000 is a powerful processor providing several key features that could make the obstacle detection unit a special application for Fukushin’s POLCAR.

California-based Syntiant was in the news with their Neural Decision Processor, the new NDP200. Syntiant has designed this processor for applications using ultra-low-power, especially useful for deep learning. With a copyrighted Syntiant core as its core architecture, it has an embedded processor, the ARM Cortex-M0. With this combination, the NDP200 achieves operating speeds up to 100 MHz.

Meant for running deep neural networks like RNNs and CNNs, Syntiant has optimized the NDP200, especially for power efficiency. Deep neural networks are necessary for computer vision applications.

Syntiant claims NDP200 performs vision processing at high inference accuracies. It does this while keeping the power consumption below 1 mW. Judging its performance, the chip could reach an inference acceleration of more than 6.4 GOP per second, while supporting more than 7 million parameters. This makes the NDP200 suitable for edge computing of larger networks.

Syntiant expects its chip will be suitable for battery-powered vision applications, such as security cameras and doorbells. In fact, the combination of the chip’s capability to run deep neural networks and power efficiency can allow it to take the next evolutionary step towards creating a better processor for computer vision applications.

Redefining computer vision

Google’s Tango prototype is a handset that can map 3D spaces simply with a walkthrough. That is possible because at its heart is the Movidius Myriad 1 vision processing unit or the VPU. According to Movidius, this VPU (not to be confused with video processing unit), is about ten times faster and has very little resemblance to GPUs or graphics processing units with which we are all familiar.

The VPU actually sits between the camera and the application processor in contrast to the GPU, which resides between the application processor and the display. However, that is only the beginning of their differences, since, as defined by Movidius, the VPU is an essential new component that will bring about astonishing changes to visual awareness in a camera.

The CEO of Movidius, Remi El-Quazzone, believes that all cameras, specifically the mobile ones are currently passing through a revolution and he calls this computational imaging, bringing in new functionality. He further explains that Movidius is developing visual processing units with functions similar to that of the visual cortex of the brain. The aim is to let the devices have the same kind of awareness and realism that the eye-brain combination has in the human body.

If you look closely at the graphical processing units, most are mere bit-bangers. These are vector processors performing identical operations on each pixel on the screen at extraordinarily high speeds. On the other hand, the VPU first interprets the data coming from the camera – very similar to what the eye and visual cortex do – before sending it to the applications processor. Therefore, instead of raw pixels, the application processor gets to work on high-level metadata, identifying where an object begins and where it ends, which ones are in front of the others, what kind of object each is, where its shadow is, the trajectory it is following, and other similar dozens of smart information. In fact, not only does it make the work of the applications processor markedly easier, it also makes possible Nuevo applications that no-one could have thought of earlier.

According to Remi, the Movidius methodology is to convert all the photons captured by the camera into metadata that expresses an understanding of the scene. Depending on the application, this metadata could then be used in a number of different ways. However, initially, they are looking into providing total visual awareness of the most relevant details in the scene.

Others have already explored the algorithms required by Movidius to perform such types of analyzes. They find that this requires supercomputers consuming megawatts to do the same. However, Movidius boldly claims that it is possible to equal or even exceed the visual awareness of such applications, consuming only few watts of power, and sometimes only a fraction of a watt.

Movidius claims a novel micro-architecture of cores entirely optimized for computational imaging. This involves structuring the delays between stages and an extremely innovative fabric of memory that allows maximizing the data localization. Since this drastically reduces the need for external memory accesses, it also reduces power requirements substantially.