Embedded

Swiftly following the announcement of the Kria K26 System on module (SOM) and KV260 Vision AI starter kit by Xilinx, Pinnacle Imaging Systems has introduced the Denali 3.0 soft image signal processor and high dynamic range (HDR) sensor module for use with the SOM and kit.

The KV260 Vision AI starter kit was created to help artificial intelligence (AI) innovators streamline development of smart vision applications and rapidly deploy them in production with the Kria K26 SOM. Pinnacle Imaging’s real time HDR imaging capability (up to 120dB) will enable AI developers to capture data-rich imaging streams for neural networks even in high contrast lighting conditions, says the company.

Pinnacle Imaging is also announcing an imager access system (IAS) HDR sensor module, paired with and tuned for the KV260 Vison AI starter kit. Th sensor module is based on ON Semiconductor’s AR0239 CMOS imager which can capture up to 120dB/20EV of dynamic range at 1080p/30 full HD resolution. The sensor captures three exposures per frame which are then merged and tone mapped in the ISP for HDR video optimised for AI-enabled security and surveillance applications.

“Having the powerful capabilities of Pinnacle Imaging’s Denali Soft HDR ISP available on the Kria KV260 starter kit will be invaluable for designers developing next generation vison systems designed to work in variable or difficult lighting conditions.” Chetan Khona, director of industrial, vision, healthcare and sciences at Xilinx.

The Denali 3.0 programmable ISP IP leverages Pinnacle Imaging’s proprietary algorithms to accurately tone map high contrast scenes and extract the full scope of rich image data for mission critical computer vision and AI applications requiring real-time imaging.

“Xilinx is really pushing to democratise these development tools and make them available to a wider audience of innovative thinkers,” said Alfred Zee, CEO of Pinnacle Imaging Systems. “The beauty of our programmable ISP on the Xilinx platform, is that it is easy for us to customise it to meet specific customers’ needs or adapt it to any sensor type, even those with non-traditional colour filter arrays. The AR0239 based IAS HDR sensor module is our first sensor module for this platform, but we expect to support additional sensors in the future based on customer requirements.”

Denali 3.0 delivers real-time video capable of producing up to 120dB or 20-EV steps of dynamic range while ensuring low latency. The HDR IP eliminates halo artefacts and colour shifts to capture up to 1080p HDR video in real time that is fully tone-mapped at 30 frames per second. Denali 3.0 is also able to accommodate automated or full-user control modes.

The Denali 3.0 ISP will be available as a trial download with licensed IP from the Xilinx app store connected to the Kria KV260 Vision AI starter kit from 01 June 2021. Pinnacle Imaging’s AR0239 HDR Sensor module will be available for sale from Avnet.

https://pinnacleimagingsystems.com/

Two DSP IP cores for embedded vision and artificial intelligence (AI) have been added to Cadence Design System’s Tensilica Vision DSP family.

The Vision Q8 and Vision P1 DSPs are designed for automotive, mobile and consumer markets. The Cadence Tensilica Vision Q8 DSP delivers what is claimed to be an industry-leading 3.8T operations per second (TOPS), to double performance and memory bandwidth compared to the Tensilica Vision Q7 DSP. It also increases energy efficiency for high-end vision and imaging applications in the automotive and mobile markets. The Tensilica Vision P1 DSP is an energy-efficient DSP, optimised for always-on and smart sensor applications in the consumer market.

Based on the SIMD and VLIW architecture found in the existing Tensilica Vision DSPs, the Vision Q8 and Vision P1 DSPs feature an N-way programming model that preserves software compatibility for an easy migration from earlier Tensilica Vision DSPs with different SIMD widths. Like the rest of the Tensilica Vision DSP family, the Vision Q8 and Vision P1 DSPs support Tensilica Instruction Extension (TIE) language, allowing customers to customise the instruction set. Both DSPs also support Xtensa Neural Network Compiler (XNNC) and the Android Neural Networks API (NNAPI) for neural network support. In addition, they support more than 1,700 OpenCV-based vision library functions, OpenCL and the Halide compiler for computer vision and imaging applications. Both cores are automotive-ready with ASIL B hardware random faults and ASIL D systematic fault certification.

The seventh-generation Tensilica Vision Q8 DSP is optimised for mobile and multi-camera automotive applications. The single core simplifies system design, reducing power by up to 20 per cent, claims Cadence while an up to four-fold improvement in performance for non-convolution layers addresses the AI workload.

The Tensilica Vision P1 DSP is optimised for always-on applications including smart sensors, AR/VR glasses and IoT/smart home devices. The 128-bit SIMD architecture delivers 400G operations per second (GOPS) and offers one third the power and area, compared to the Vision P6 DSP, yet with 20 per cent higher frequency, says Cadence. The architecture is optimised for small memory footprint and operation in low power mode.

The Tensilica Vision Q8 and Vision P1 DSPs support Cadence’s Intelligent System Design strategy. The Tensilica Vision Q8 DSP is available now, while the Tensilica Vision P1 DSP is expected to be available for general release in the second quarter of 2021.

http://www.cadence.com