Periodic Reporting for period 2 - Hailo-8 (End-to-end hardware implementation of Artificial Neural Networks for Edge Computing in Autonomous Vehicles)
Période du rapport: 2019-12-01 au 2020-08-31
The project is targeting a completely new architecture for computers that will be able to run full-scale neural networks on devices that are not residing in data centers and thus address heavy-duty machine learning workloads. The main innovation is around the core architecture that is completely different from traditional computers.
This is important to unleash the possibilities of introducing the power of machine learning to devices that were not able to consider this before like cameras, personal assistance devices, robotics etc.
Among other segments, Hailo Technology will be able to address the automotive industry needs in growing autonomy levels by providing better perception capabilities within heat dissipation limitations that are required by vehicles. As a result, compliance with safety regulations may be increased.
This is important to unleash the possibilities of introducing the power of machine learning to devices that were not able to consider this before like cameras, personal assistance devices, robotics etc.
Among other segments, Hailo Technology will be able to address the automotive industry needs in growing autonomy levels by providing better perception capabilities within heat dissipation limitations that are required by vehicles. As a result, compliance with safety regulations may be increased.
The work that was performed during this phase included activities in the following domains:
• Planning for design testability and more specific safety aware mechanisms – this involved training, engaging with domain experts and ramping up all the Quality Assurance (QA) related activity to guarantee compliance with regulatory requirements
• Planning the post-production activity-related activities – this included engaging with relevant providers that will conduct the procedure once the device is manufactured. Introduction of all the production line related functions to the silicon and planning the test procedures for the device.
• Fast track program launched and executed allowing to onboard over a dozen customers in various stages of evaluation. These were done with the Evaluation Board (EVB) platform that was developed and manufactured for the purpose. Successfully executing this activity puts a high likelihood to identify high-quality potential customers and partners which are targeted by the device once available and become customers.
• Ramping a new website and other marketing activities to promote the brand and the accessibility to target markets. Resulting in growing traffic and rapid expansion of market reach and market segment diversity.
• Completing all activities related to mass production readiness which includes the test plan completion for all stages of post-production as well as the load boards that are needed to run the device lots through the test equipment
• Successful fabrication and bring-up of the production device samples and validation that it reaches (and exceeds) expected performance
• Developing and manufacturing a set of deployment platforms in standard form factor ready for deployment in standard compute platforms
• Preparing the marketing channels for promoting the product
• Active business development work in targeted market segments to increase product demand
• Planning for design testability and more specific safety aware mechanisms – this involved training, engaging with domain experts and ramping up all the Quality Assurance (QA) related activity to guarantee compliance with regulatory requirements
• Planning the post-production activity-related activities – this included engaging with relevant providers that will conduct the procedure once the device is manufactured. Introduction of all the production line related functions to the silicon and planning the test procedures for the device.
• Fast track program launched and executed allowing to onboard over a dozen customers in various stages of evaluation. These were done with the Evaluation Board (EVB) platform that was developed and manufactured for the purpose. Successfully executing this activity puts a high likelihood to identify high-quality potential customers and partners which are targeted by the device once available and become customers.
• Ramping a new website and other marketing activities to promote the brand and the accessibility to target markets. Resulting in growing traffic and rapid expansion of market reach and market segment diversity.
• Completing all activities related to mass production readiness which includes the test plan completion for all stages of post-production as well as the load boards that are needed to run the device lots through the test equipment
• Successful fabrication and bring-up of the production device samples and validation that it reaches (and exceeds) expected performance
• Developing and manufacturing a set of deployment platforms in standard form factor ready for deployment in standard compute platforms
• Preparing the marketing channels for promoting the product
• Active business development work in targeted market segments to increase product demand
The Hailo-8 device represents a disruption in the edge processing industry by enabling access to advanced machine learning capabilities to entry-level and mid-range products outside of data centers. Traction from lead markets is evident and allows us to fine-tune the offering accordingly and design next-generation products to widen the product portfolio further and widen our market position. In the course of the project and in the period that follows we see a growing funnel of potential customers that are already in evaluation or planning to evaluate the Hailo-8 towards inclusion in their final product. This will hopefully be a benefit for huge ecosystems like automotive safety, smart cities, industrial automation and more. Furthermore, from a socio-economic perspective, this device is focused on energy efficiency and allows to extract more from the energy it uses and thus, in the longer term, lower overall energy consumption of large systems, lower accidents and improve quality of life in related systems that incorporate it.