Today, Hailo, an AI chipmaker based in Tel Aviv with a lot of capital, released the newest version of their processor family, the Hailo-15H, M and L SoCs. Just like the Hailo-8, these new processors enable AI-supported technology at the border, mainly to promote smart camera performances.
According to Orr Danon, the Chief Executive Officer of Hailo, Hailo-15 is a major advancement for making AI accessible and cost-friendly at the edge. Thanks to their edge solutions that have already been used by hundreds of clients around the globe, the accomplishment of their AI technology, and their batch of software, they are able to offer AI in a camera design.
The business claims the Hailo family will have operational paces extending from 7 TOPS to 20 TOPS, making it 5x faster than similar priced solutions. With the Hailo-15H, users can run a model like YOLOv5 in real time at 1280×1280 resolution or ResNet-50 at 700 frames per second. If you don’t need that much real-time operation, or are okay with fewer frames per second, the processor can handle 4k60 video input.
This business emphasizes the utilization of vision-based technology to allow cameras to rapidly detect objects. Additionally, their chips can be employed for video enhancement, for example, improving video in inferior lighting settings, video steadying, and High Dynamic Range imagery.
Hailo-15 enabled cameras allow users to benefit from greater savings, better privacy protection and faster video analysis in cases with large numbers of cameras. By employing cloud analytics, it reduces the use of video bandwidth and data processing, while anonymizing data at the source. This allows for a high quality, artificial intelligence-backed video monitoring solution that increases people’s safety, protecting their privacy and enabling businesses to carry out their operations faster and more efficiently. At the same time, its installation costs and networking infrastructure requirements are minimized.
Hailo is now taking its existing groundbreaking architecture to the next level with its new processors. This design helps the company to analyze the neural network being processed on the chip and then effectively use the physical resources available, making sure the compute, memory and control systems are as close to each other as possible to create maximum efficiency. Additionally, the production also includes the ability to shut off sections of the chip that are not required for current operations.