March 15, 2022
Began to manufacture the chip on an industrial standard. Has signed a partnership agreement with CVEDIA that brings it to the car, industry and security thermal camera market. CEO Or Danon to Techtime: “What you see today in the cloud – will reach the end fittings tomorrow”
Pictured above: Or Danon with Halo’s AI processor. Startup chips worth more than $ 1 billion
The Tel Aviv company Hailo is adapting its artificial intelligence processor, Hailo-8, to the demanding market of artificial intelligence with critical end accessories, such as industrial, military systems and cameras for ADAS applications in the automotive industry. In a conversation with Techtime, the company’s CEO, Or Danon, said that this is a market that is growing. “Recently it became clear to us that there is a need for support for extreme temperatures, and in recent weeks we have launched a new version of our modules that supports temperatures defined by industry standards: between minus 40 ° C and plus 85 ° C. The update was released after we received a demand from customers of industrial products and customers in the automotive field. “
A synthetic IR image created on a computer rather than a camera
As part of this strategy, the company announced today (Tuesday) an agreement for cooperation with a company CVEDIA British, which develops image processing applications. The collaboration focuses on adapting CVEDIA’s thermal camera software to run smoothly on Halo processors. This is an unusual solution: Because there is a great shortage of IR image databases, CVEDIA has developed a solution based on an algorithm that produces synthetic images: instead of training the neural network using images of objects collected from huge databases, the algorithm produces the information itself.
It builds a three-dimensional model of the object to be studied (ship, plane, vehicle, human, animal, etc.), produces millions of artificial images of it on different computers and in different environmental conditions, and feeds these images into the neural network to train it. The collaboration will open up for Halo the opportunity to reach CVEDIA customers who use thermal cameras for applications such as security and automotive systems (ADAS). These are highly coveted customers like a company FLIR Systems (Owned by Teledyne), which manufactures thermal cameras for applications, security, industry, automotive and aviation.
Halo is growing with the smart end fittings market
Danon: “The field of artificial intelligence in end fittings includes a lot of innovative products that are coming to market. Nowadays for example, AI props are also concerned with image enhancement and not just image analysis. This development illustrates our central perception that everything we see today in the cloud will reach the edge fittings tomorrow. We feel there is recognition that we are providing an exceptional solution in the field. Chipmakers like NXP treat us as an important partner. Together with our chip, we also provide them with a complete package of software solutions – from training, deployment, working with different CPU architectures to working on different operating systems. “
“Today there are about 150 companies developing products based on our chips. We already employ 180 employees, 160 of them in Israel and the rest in our offices around the world. ” Hailo was founded in February 2017 by Or Danon, Avi Baum, Hadar Zeitlin and the late Rami Feig, and developed a neural network processor (Deep Learning) for running artificial intelligence applications with end accessories. The company’s technology is based on an interesting phenomenon that the company has identified: in the process of processing the inferences in a neural network, there are differences between the behavior of the different layers in the neural network, and therefore they need different resources.
The company’s chip knows how to analyze the specific task, map out the resources needed for each processing layer, and provide them as needed. In this way it speeds up the processing process and reduces power consumption. The company’s flagship chip, the Hailo-8, manages to provide a processing power of 26 trillion processing operations per second (26TOPS) with a typical power consumption of only about 2.5 watts. In recent months the company has reported a long line of distribution agreements and planning victories. This has not gone unnoticed by investors – in October 2021, it raised capital of about $ 136 million at a company value of about $ 1.13 billion after the money.
Posted in categories: Artificial Intelligence, News, Semiconductors, Israeli Industry
Posted in tags: AI, Artificial Intelligence, Halo, Chips
