Hailo’s AI hardware devices present design engineers with a lowest-cost-per-TOPS solution for integrating edge-AI acceleration in conjunction with standard host processors. This user-defined board page serves as an overview of Hailo’s product range.
By filling out the form below with details of your commercial project, you can evaluate Hailo’s technology through ipXchange. Please specify the product family you wish to test if known.
Hailo-8 AI Accelerator
With 26-TOPS (Tera-Operations-Per-Second) performance and an average power consumption of 2.5 W, Hailo-8 is perfect for real-time edge AI inferencing at a level previously unseen at this size and price on the market. The Hailo-8 chip itself is smaller than a penny, and Hailo presents this solution on various M.2 card options for easy integration into your design without the requirement for additional external memory.
The Hailo-8’s neural networking architecture enables designers building processor-driven devices to run deep learning applications at far lower power, cost, and latency than traditional AI solutions. It can also perform simultaneous processing of multiple workloads and supports the use of multiple Hailo-8 devices for the most demanding tasts.
Hailo-8 is available in industrial- and automotive-grade models and supports TensorFlow, TensorFlow Lite, Keras, PyTorch, and ONNX AI frameworks, X86 and Arm host architectures, and Linux and Windows operating systems.
Hailo-15 AI Vision Processor
The Hailo-15 series presents designers with an AI hardware accelerator made specifically with camera, image, and video processing workloads in mind. These Systems-on-Chip (SoCs) provide top-tier real-time AI inferencing – such as model detection – image enhancement, and other deep-learning AI algorithms while achieving the same power consumption as a typical camera SoC.
In addition to Hailo’s AI neural network core, the Hailo-15 series supports high-frame-rate processing and supports 4K60 image quality with the built-in High Dynamic Range (HDR) and Noise Reduction (NR) algorithms that one would expect from a top-tier camera, with the DSP and video encoding to match.
The industrial-grade Hailo-15 SoCs are available in options for 20, 11, and 7 TOPS of processing capability. Hailo has created a Yocto Linux SBC development kit for easy testing of these devices with up to 2 image sensors provided by the user.
Hailo-10H M.2 Generative AI Acceleration Module
The Hailo-10H acceleration module puts 40 TOPS of generative AI power into an M.2 form factor for workloads that include Large Language Models (LLMs) to be performed at the edge with minimal strain on the host CPU/GPU.
With a typical power consumption of 3.5 W, this is a remarkable offering that enables low-latency, real-time deep-learning and generative AI in consumer PCs, smart cars, and even IoT devices for natural language conversations without the requirement for an internet connection.
As with the Hailo-8, multiple workloads can be run on a single device, and the Hailo-10H supports simultaneous operation of multiple units in parallel.
The M.2 form factor makes it extremely easy to integrate and test the Hailo-10H, and with Hailo’s vast range of supporting software and development suites for AI workloads, you can get started developing in no time.
As previously noted, fill out the form below to apply to evaluate any of Hailo’s technologies for use in a commercial project, and ipXchange will get you connected for more information and a sample or development kit as required.
