Why the chip in this new drone is good news for smartphones
You probably haven’t heard of Movidius or its processors, but the European startup has just announced its new Myriad 2 vision processor that could set the company up as a preponderant player in the mobile chip market in the future.
The Myriad 2 isn’t going to replace your main smartphone processor from the likes of Qualcomm, MediaTek or Samsung. Instead, it’s a highly specialized complementary chip designed to perform complex tasks in the field of image and video processing. This is an important capability for a wide range of products, from security cameras to drones, and even novel pieces of kit like Google’s Project Tango.
In fact, the Myriad 1 processor was the chip that helped to power Google’s first generation Project Tango tablets. More recently, the Myriad 2 MA2100 is the chip behind DJI’s latest drone, the Phantom 4. This drone offers up impressive obstacle detection systems based on computer vision, at a consumer friendly price that won’t break the bank. You can check out the Phantom 4 drone in action along with a little explanation about the technology provided by Movidius in this video:
Movidius is positioning its specialized Visual Processing Unit as a specialized alternative to today’s common processors for visual processing. The Myriad 2 a custom built piece of silicon that features a set of camera interfaces, enhanced vision accelerators, a group of 12 specialized vector VLIW processors called SHAVEs, and an “intelligent memory fabric” to enable power efficient processing. The processor also has a very small area footprint, measuring just 35 square millimetres, which makes it relatively cost effective to implement.
If that technical jargon makes you a tad somnolent, essentially this specialized chip is faster and more efficient at vision processing tasks than your regular CPU or GPU.
Why does this particularly matter? Well, it has very useful practical applications not just for drones, but for growing markets such as virtual reality and visual recognition computer learning. Take the new VR headsets from HTC, Oculus, and Sony, which make use extra tracking components, such as laser light boxes, to keep up with a user’s movement. However, this could perhaps be simplified with Project Tango-eqsue cameras actually on the product itself, providing that you have an efficient and powerful enough processor to interpret depth and distance data.
Furthermore, Google is said to be sourcing a MA2150 version of this processor for an undisclosed product. A smartphone version of Project Tango is expected to arrive this summer and the two announced a partnership back in January to work together on deep learning, with image recognition algorithms actually running on a device in real-time. This could open up new areas for superior object, place, and product recognition using a smartphone’s camera.
For a closer look at what Movidius plans to bring to the mobile market and how its latest VPU works, have a watch of the video below.