Next-generation thermal edge AI solutions

1 min read

Hailo, an artificial intelligence chipmaker, has partnered with CVEDIA, a supplier of edge computer vision software, in launching next-generation thermal edge AI solutions.

These joint solutions combine CVEDIA’s computer vision expertise and synthetic data technology with the Hailo-8 AI processor, offering customers high-performance, scalable thermal edge AI solutions for a wide range of market segments including industry 4.0, smart cities, security, and transportation.

Thermal imaging sensors are employed in security, surveillance, and Advanced Driver Assistance Systems (ADAS) applications due to their ability to operate in bad weather, low light conditions, and at night time, while providing fewer false positives in such challenging environments. 

Developing thermal AI solutions at scale, however, can be challenging given the scarcity of thermal training data publicly available. CVEDIA’s synthetic data pipeline addresses common data bottlenecks including data collection, annotation, and validation, while enabling rapid scalability and maintenance of thermal camera-based solutions.

The Hailo-8 AI processor, with up to 26 tera-operations per second (TOPS) at a typical power consumption of 2.5W, improves the performance of object detection and classification algorithms. The company’s joint solutions will make AI models significantly smaller, enabling them to run faster without undermining their accuracy, while helping to solve pain points such as speed, dependability, time to market, and production costs.

“CVEDIA is an end-to-end computer vision solutions company that enables companies to rapidly build, scale and maintain their AI products across the commercial and defence sectors,” said Arjan Wijnveen, CEO at CVEDIA. “When combined with Hailo’s high-performance AI processor, we can provide incredibly power-efficient, highly scalable, and safer AI solutions for a new era of intelligent cameras.”

Commenting Liran Bar, VP Business Development at Hailo said, “This partnership bolsters our position in the video analytics and security sector, enabling us to further address the rapidly growing need for efficient and reliable technologies that process visual images more efficiently at greater speeds, while still maintaining data privacy.”

The Hailo-8 processor has been built with an architecture that enables edge devices to run sophisticated deep learning applications that could previously only run on the cloud.

CVEDIA’s thermal perimeter security solution is running at 55 Frames Per Second (FPS) on the Hailo-8 AI processor, providing over a 20 times faster FPS rate when compared to many GPU-based solutions, with a latency of 18.5 ms, power consumption of only 1.8W, and minimal accuracy loss.