Qualcomm Makes First Purpose-Built Chips For Smart Cameras

Matt Broersma is a long standing tech freelance, who has worked for Ziff-Davis, ZDnet and other leading publications

Follow on: Google +

The new Vision Intelligence Platform features dedicated hardware for image capture, as well as built-in AI

Qualcomm has introduced its first purpose-built chips for internet-connected devices, focusing on image processing and built-in artificial intelligence to support it.

The new Vision Intelligence Platform follows a broader trend that has seen companies such as Huawei add AI capabilities to its chips for mobile phones, minimising the data that has to be processed remotely. ARM last month also said it would build Nvidia AI capabilities into its widely used mobile chips.

Qualcomm’s new platform means embedded cameras in areas such as sports, security or robotics can access advanced features such as extreme low-light processing, image stabilisation and obstacle avoidance.

Particular applications could include helmet-mounted sports cams and domestic robots that need to navigate around a house, Qualcomm said.

IoT cameras

Qualcomm has made chips for such so-called Internet of Things (IoT) devices in the past, but has previously used repurposed smartphone chips.

The new line uses dedicated hardware for chores that might not be needed on a smartphone, meaning the camera features can be carried out using less power than would be the case with a general-purpose chip.

The first chips in the line, the QCS605 and QCS603, both include onboard Wi-Fi. The 605 supports up to simultaneous 4K and 1080p video streams at 60 fps, while the 603 supports 4K and 720p streams at 30 fps. Both can handle a greater number of streams at lower resolutions.

The 603 runs at lower power and has the smaller footprint of the two, with quad-core processing and 1×1 802.11ac Wi-Fi. The 605 has higher specifications, including 8 cores and 2×2 802.11ac Wi-Fi.

The platform includes Qualcomm’s Snapdragon Neural Processing Engine (NPE) software framework, which is compatible with AI libraries such as Tensorflow, Caffe and Caffe2, the Android Neural Networks API and Qualcomm’s Hexagon Neural Network.

It also includes software development kits (SDKs) for Qualcomm’s camera processing, machine learning and computer vision software.

Qualcomm has already begun sampling the 605 and 603 to hardware partners. The company said reference designs for a 605-based virtual reality camera is available now, with a 603-based industrial security camera reference design available in the second half of 2018.

Put your knowledge of artificial intelligence (AI) to the test. Try our quiz!

Read also :