Swiss startup launches mega-neuron vision processor

April 15, 2019 //By Peter Clarke
Swiss startup launches mega-neuron vision processor
aiCTX AG (Zurich, Switzerland) has announced it has created a fully-asynchronous, event-driven neuromorphic processor – the DynapCNN – for always-on, real-time dynamic vision applications.

The company is a China-backed startup spun out of the Institute of Neuroinformatics at the University of Zurich in March 2017.

In January 2018 the company announced a $1.2 million seed funding round, led by Pre Angel Capital (Zhuji Jiawei). In November 2018 the company added to that $1.5 million from Baidu Ventures.

The DynapCNN is a 12 square millimeter chip fabricated in 22nm manufacturing process technology that houses overs 1 million spiking neurons suited for implementing convolutional neural networks. The company claims that the DynapCNN is the most power-efficient way of processing data generated by event-based and dynamic vision sensors.

The company also claims that the DynapCNN is 100 to 1,000 time more power efficient than other "state-of-the-art" approaches and delivers 10 times shorter latencies in real-time vision processing. This, in turn, opens up the prospect of long-life battery-operated equipment.

Computation in the DynapCNN is triggered directly by changes in the visual scene, without using a high-speed clock or frames, showing similarities to the approach used by image sensors from French startup Prophesee (see Chronocam changes name, raises $19 million ) and subsequently discussed by Qualcomm.

aiCTX claims its continuous computation enables latencies of less than 5ms and at least a 10x improvement over current deep learning solutions for real-time vision processing.

Development kits for the DynapCNN processor will be available in 3Q19.

Related links and articles:

www.aictx.ai

News articles:

Development board offered for event-based vision

Chronocam changes name, raises $19 million

Eta adds spiking neural network support to MCU

BrainChip launches spiking neural network SoC


Vous êtes certain ?

Si vous désactivez les cookies, vous ne pouvez plus naviguer sur le site.

Vous allez être rediriger vers Google.