On-device computing solutions startup Perceive emerged from stealth today with its first product: the Ergo edge processor for AI inference. CEO Steve Teig claims the chip, which is designed for consumer devices like security cameras, connected appliances, and mobile phones, delivers “breakthrough” accuracy and performance in its class.
“We have been working with industry leaders such as Arlo since our inception and look forward to supporting them as they build amazing products that take advantage of the capabilities of Ergo,” Perceive VP of marketing David McIntyre told VentureBeat via email. “We look forward to partnering with Arlo to reinforce our shared focus on privacy and customer-centric innovation.”
By eliminating the need to send data to the cloud for analysis, Ergo could bolster battery life while providing peace of mind to homeowners and businesses about privacy. In fact, Perceive says it has already partnered with a smart home security camera brand to integrate machine learning applications into the manufacturer’s future products, lending legitimacy to the startup in a crowded market.
Ergo delivers around 4 theoretical operations per second (TOPS), with the ability to run AI and machine learning models in a range of architectures simultaneously, powering applications like audio event detection and speech recognition. For example, Ergo can run YOLOv3, a popular object detection algorithm, for up to 246 frames at 30 frames per second while consuming about 20 milliwatts of power.
Ergo requires no external RAM, and its 7 x 7-millimeter package makes it well suited for use in phones, cameras, and other small electronics, Perceive says. Theoretically, it’s able to achieve 55 TOPS/watt, which is roughly 20 to 100 times that of leading edge processors.
Here’s how that compares:
- Hailo-8: 26 TOPS (2.8 TOPS per watt)
- Nvidia Jetson Xavier NX: 21 TOPS (1.4 TOPS per watt)
- Google’s Edge TPU: 4 TOPS (2 TOPS per watt)
- AIStorm: 2.5 TOPS (10 TOPS per watt)
- Kneron KL520: 0.3 TOPS (1.5 TOPS per watt)
Along with Ergo, Perceive says it will provide a “complete solution” to OEMs, including reference boards as well as standard imaging and audio inferencing applications for common inferencing tasks. Customers will also be able to tune their applications or create novel applications with support from Perceive.
Perceive, which was founded in 2018, was incubated by and is a majority-owned subsidiary of Xperi, a San Jose, California-based firm that licenses technology and intellectual property in areas such as mobile computing, communications, memory, and data storage. Xperi is perhaps best known for brands like DTS, IMAX Enhanced, and HD Radio, which hundreds of partners currently use in billions of products.
It’s worth noting that Perceive has plenty in the way of competition. Startups Hailo, AIStorm, Esperanto Technologies, Quadric, Graphcore, Xnor, and Flex Logix are developing chips customized for AI workloads — and they’re far from the only ones. The machine learning chip segment was valued at $6.6 billion in 2018, according to Allied Market Research, and it is projected to reach $91.1 billion by 2025.
Mobileye, the Tel Aviv company Intel acquired for $15.3 billion in March 2017, offers a computer vision processing solution for AVs in its EyeQ product line. Baidu in July unveiled Kunlun, a chip for edge computing on devices and in the cloud via datacenters. Chinese retail giant Alibaba said it launched an AI inference chip for autonomous driving, smart cities, and logistics verticals in the second half of 2019. And looming on the horizon is Intel’s Nervana, a chip optimized for image recognition that can distribute neural network parameters across multiple chips, achieving very high parallelism.
Author: Kyle Wiggers.
Source: Venturebeat
