Introducing the first realtime AI enabled imaging engine with support for encrypted networks, our Image Processing Unit (IPU) looks toward the future of real time camera applications where AI is on the edge. Our solution delivers latency, lower bandwidth and substantially reduced power consumption.
FotoNation’s third generation IPU is a collection of high performance post silicon configurable IP cores optimized to resolve a core type of image processing task that provide ‘unfair’ advantage for our high quality imaging solutions.
IPU showcases more than a decade of experience in hardware design and hybrid deployment. It brings together signal processing and sensor fusion for optimum analytics. It has been designed with deep learning-based imaging solutions to deliver better user experiences by understanding, enhancing and personalizing your world.
Imaging AI enabled IP core (engine) that runs in real-time.
Offers support for encrypted neural networks.
Privacy, trust and authentication in the context of edge computing.
Enables Always-On inference (operates even when the device is “off”).
Configurability post silicon via programmable inference engines.
Unmatched performance and power consumption.
The future of real time camera applications using AI is on the edge where latency, lower bandwidth and substantially reduced power consumption is key to delivering always ON applications.
IPU is guaranteed to run multiple concurrent high performance image enhancement and analytics solutions from FotoNation.
Building AI centric solutions for camera applications requires a holistic approach. FotoNation IPU brings together signal processing and sensor fusion for optimum analytics, deep learning based imaging solutions that deliver better user experience by understanding, enhancing and personalizing your world.
Based on Xperi’s DBi created to address memory bandwidth – deep learning’s biggest constraint and a must for all low power smart imaging applications – IPU also enables unique chip architectures.
At the core of IPU is a deep learning imaging infrastructure with billions of ground truth data sets for training and validation and a solid testing framework with tens of millions of marked real world images for verification and testing.
Depending on customer and target market, IPU can be configured to be as simple as a dedicated engine (e.g. FDE – Face Detection Engine) or as complex as a custom solution supporting multiple imaging and vision solutions.
Deliverables include encrypted RTL, device drivers and software binaries as well as PC development tools.