Hailo Democratizes Edge Artificial Intelligence Through Power Efficient Edge Computing Box

Subscribe To Download This Insight

By Lian Jye Su | 3Q 2021 | IN-6233

Hailo is solidifying its position as a strong contender in the edge AI space with partnernships and extensive ML toolkits.

Registered users can unlock up to five pieces of premium content each month.

Log in or register to unlock this Insight.

 

AI in the Box

NEWS


In June 2021, Hailo, an Israel-based edge Artificial Intelligence (AI) chipset vendor, announced its partnerships with several Taiwan-based industrial equipment manufacturers, including Lanner Electronics and Vecom. These partnerships involve the inclusion of Hailo-8, Hailo’s Machine Learning (ML) inference chipset, in industrial-grade fanless x86 computing boxes. These boxes can support multiple cameras through digital input-output (I/Os), multiple serial communications, and USB ports, along with gigabit network interfaces. In addition, they are relatively small in terms of footprint, designed for outdoor usage through wall-mount installation.

These announcements came on top of Hailo’s partnership with Foxconn and Socionext in the launch of BOXiedge, an edge computing box for ML-based machine vision, in May 2020. Hailo envisions these edge computing boxes being widely adopted in smart city, smart retail, and manufacturing applications. These edge boxes can support high-performing ML inference engines for mission- and business-critical machine vision applications, including video analytics, traffic management, and access control in a power-efficient manner.

AI-Enabled IoT Gateway Potential Gamechanger

IMPACT


With the introduction of ML capability, these edge computing boxes can provide a seamless transition of legacy equipment into a large fleet of intelligent IoT devices. Previously, legacy devices in the field had to rely on constant cloud connectivity for cloud-based analytics, incurring massive connectivity costs and security concerns. In addition, these devices are designed to work for an extended period and can be extremely costly to be replaced. By connecting them to an edge computing box with localized ML processing, enterprises can introduce localized intelligence to their legacy equipment, reduce cybersecurity risk exposure and connectivity cost, and accelerate the convergence of Information Technologies (IT) and Operational Technologies (OT).

More important, enterprises are starting to appreciate the value of these gateways. According to the report on Smart IoT Gateways, ABI Research estimates that a total of 21.4 million next-generation smart IoT gateways shipped in 2025. Close to 60% of these gateways will have ML inference or training chipsets. In addition to ML capability, these gateways also enable increased visibility, monitoring, and management of IoT devices and enhanced security by including hardware-based security solutions.

Democratize Edge AI Through Hardware and Software

RECOMMENDATIONS


In recent years, edge computing boxes have become a target market for many major AI chipset vendors. For example, NVIDIA has recently launched Jetson AGX Xavier Industrial Module. Intel and Qualcomm are also very active in this space, leaning on their respective edge ML inference chipset products, Movidius Myriad X and Cloud AI 100. Hailo is not shy to point out that it competes directly against Intel’s Movidius Myriad X and Google’s Edge TPU. Thus far, it has demonstrated its strength in high inference performance without sacrificing power efficiency. However, to remain competitive, partnership with prominent industrial hardware vendors is not sufficient.

As such, Hailo has significantly boosted its ML developer toolkit. Starting as a chipset company, Hailo has been aggressively developing its software capabilities. Developers working on Hailo-8 can leverage the Hailo Software Toolchain, consisting of a dataflow compiler, runtime software, and application toolkit. Hailo also offers pre-trained models for specific applications, such as classification, depth and pose estimation, lane detection, facial detection and recognition, and semantic segmentation.

Developer-friendly software and toolkits are critical in democratizing edge AI. Unfortunately, data science and ML talent are scarce in the industrial sector. As a result, most edge AI chipset vendors have been actively offering comprehensive developer-focused software solutions, ranging from established vendors like NVIDIA, Intel, Qualcomm, and NXP, to startups like Blaize, Gyrfalcon Technology, and Syntiant. With this offering, Hailo is solidifying its position as a strong contender in the edge AI space.