Intel video AI computing box: integrating software and hardware with AI algorithm to accelerate intelligent video analysis in thousands of industries.
Artificial intelligence is moving from a technical concept to a thousand industries, realizing the actual landing of multiple scenes. In this process, the edge architecture design, which puts computing power closer to the data source or service delivery position, has also attracted much attention in the process of commercialization of various AI application solutions with more agile processing power, lower delay and better security. Among them, intelligent video analysis is favored by users with strong market demand and broad application scenarios, and the growth trend is strong. It is estimated that by 2022, the market scale will exceed 180 billion yuan, with a compound growth rate of 34% 1; By 2021, video data is expected to account for 82% of IP traffic. Intelligent video analysis solutions based on edge AI have been widely used in more than 10 fields such as smart city, smart community, intelligent transportation, intelligent manufacturing and smart retail.
Creating the edge AI video analysis ability with excellent performance and facing a wide range of application scenarios is inseparable from the effective aggregation of diversified algorithms and high-performance computing power. This requires not only a powerful hardware platform to carry the intensive computing load required by computer vision and reasoning, but also an edge AI video analysis capability with excellent performance and wide application scenarios, which is inseparable from the effective aggregation of diversified algorithms and high-performance computing power. This requires not only a powerful hardware platform to carry the intensive computing load required by computer vision and reasoning, but also appropriate development tools and software frameworks to complete AI acceleration, and further activate the ecology to provide solutions for more scenarios.
In order to help ISV (Independent Software Developer), ODM/OEM (Original Design Manufacturer/Original Equipment Manufacturer) and SI (System Integrator) and other partners to build highly available intelligent video analysis solutions for diverse edge scenes, Intel launched Intel Video AI Computing Box based on advanced software and hardware products and technologies such as Intel Core processor, OpenVINO™ toolkit and Intel Media SDK. While promoting the horizontal and vertical two-way ecological construction, we will join hands with partners to provide strong assistance for the intelligentization of thousands of industries.
Create embedded products with soft and hard integration
In order to help partners to complete the solution construction more efficiently and conveniently, Intel not only provides a series of advanced software and hardware products and technologies such as Intel Core processor with built-in Intel core graphics card and OpenVINO™ tool suite in the video AI computing box, but also accelerates the development and engineering landing of edge AI applications by adopting the integrated design that integrates software, hardware and algorithms.
Hardware design:
Intel Video AI Computing Box, based on a new generation of Intel Core processors, focuses on the product features of "multi-purpose with one core". In addition to the high-performance CPU as always, which provides powerful general computing power and supports workloads such as IPC access and management, device management and containerized services, the new generation of Intel Ruiju X integrated graphics card [2] provides powerful AI processing and media capabilities, which is 2.95 times higher than the previous generation.
Software architecture:
Intel provides powerful and easy-to-use software tools and software development kits to accelerate customers’ workload of building AI and video analysis based on Intel hardware platforms. As shown in the figure below, we provide partners with multi-dimensional software support such as video processing, model optimization and AI reasoning acceleration from multiple levels to help accelerate product verification, optimize performance and shorten time to market.

Intel ®Media SDK
Intel Media SDK supports hardware acceleration to realize fast video transcoding, image processing and media workflow, and provides advanced libraries, tools, examples and comprehensive and convenient API.
OpenVINOTool kit
Intel Video AI calculates and integrates OpenVINO™ tool suite, which supports customers to achieve high performance, artificial intelligence and computer vision reasoning based on hardware optimization, adapt to mainstream deep learning frameworks and networks, and accelerate the deployment of AI reasoning workload on Intel hardware platforms. It is worth mentioning that OpenVINO™ Toolkit can also help partners to speed up the development of edge AI video analysis applications with more than 40 pre-trained and optimized models and code examples provided by Open Model Zoo (covering a series of common and practical AI models such as object detection, object recognition, behavior detection and image processing).
Combine vertical and horizontal, and work together with partners to improve the edge AI video analysis ecology
With the algorithm mall (algorithm library) continuously enriched into Intel intelligent edge video analysis ecosystem, Intel video AI computing box further provides a stage for partners at different levels to reflect their own values and advantages. For example, at the product design level, it can help ODM manufacturers to quickly build a highly available edge AI video analysis platform according to the needs of end users; At the algorithm level, it can help ISV vendors to quickly customize and scene different application scenario software and algorithms, thus accelerating product definition, algorithm research and development and edge AIoT capability building; OEM and SI manufacturers can provide high-performance and high-availability overall solutions for various industries by highly integrating software and hardware and algorithms with video AI computing boxes, and accelerate the accurate landing of intelligent edge applications.

- Hardware eco-partners: a large number of ODM partners rooted in the industry application for more than ten years have been aggregated through the performance verification test plan of Intel video AI computing box, providing all kinds of video AI computing box hardware devices suitable for industry applications;
- Algorithm partner: Intel cooperates with various algorithm partners to provide cross-industry, multi-scenario and hardware-optimized AI algorithms for video AI computing boxes;
- Solution Partner: Intel cooperates with all kinds of solution partners to provide all kinds of complete solutions (hardware+algorithm+business solution) that have been deployed and verified and can be customized and expanded.
Why can Intel aibox bring stronger industry impetus?
In Intel’s view, only by adopting a way close to the real application needs can edge computing take root in the current digital society and bring real "digital intelligence" to all walks of life. Therefore, in the process of building the video AI computing box, Intel pays special attention to working with competent partners to build a multi-point collaborative and widely interconnected ecological cooperation model, give full play to the multilateral network effect brought by ecology, and work together with partners to realize the rapid evolution of the edge AI video analysis market.

Based on the conceptual design of video AI computing box, Intel, together with different types of partners, has built a series of end-to-end solutions for smart parks, retail restaurants, industrial manufacturing, medical health, smart agriculture, smart transportation and other vertical fields in dozens of scenarios such as smart payment, smart parking, smart VDD, smart whiteboard and smart conference. Focusing on the basic software and hardware capabilities provided by the video AI computing box, partners can provide end users with full-process workloads that can be delivered end-to-end, from high-density video stream coding and decoding, multi-channel display/splicing, AI reasoning, storage and cloud-side collaboration.