Computer vision and AI technologies – and video analytics specifically – are not new to enterprises. For many years businesses have heard that computer vision will be ‘the next big thing’. This hype, along with the technical challenges of implementing vision, means many remain understandably skeptical about the value of making an investment.
This is the context in which systems integrators (SIs) must approach computer vision with their customers. The key question SIs will be expected to answer is: what’s changed this time?
Artificial intelligence (AI) is generating opportunities for SIs building solutions that include applications of computer vision. AI is expected to create $13 trillion in additional economic activity by 20301. Computer vision plays an important role in generating this value, as AI-based applications demand immense quantities of data in order to deliver accurate, actionable insights. This data creates exciting opportunities for computer vision solutions, such as smart cameras at the edge, where computer vision technologies can be fully utilized.
In 2019, 2,500 petabytes of video data is generated daily just in the area of surveillance alone2. Much of that data is captured at the edge on devices like surveillance cameras and factory sensors, where AI can be used to analyze and tag the data so that businesses can act on the data in real-time. AI can also write rules or searches to help recognize particular images automatically.
AI models can also be deployed in the cloud, where they use large data sets to identify trends and patterns on which insights can be presented. Thanks to the large volumes of data now available, deep learning-based machines are now able to meet or exceed human image and speech recognition abilities.
Despite this wealth of opportunity, currently less than 10 percent of video data is viewed and analyzed2. Possible use cases for AI in computer vision include3:
- Traffic flow optimization: traffic congestion costs the U.S. $87.2 billion in wasted fuel and productivity4.
- Responsive retail: 90 percent of shoppers say they struggle to find the help they need when shopping5.
- Quality assurance: Even after nine months of training, expert human operators can only maintain 70-85 percent accuracy in identifying product defects6.
With the right technologies however, SIs can help position their customers for success. There are three critical sections to enterprise vision environments:
- Architecture: As AI continues to evolve, both machine learning and its subset, deep learning, will need highly scalable systems to cope with shifting volumes of video data, and the changing ways in which enterprises will require that data to be used. Intel’s wide range of silicon architectures optimized and purpose-built for AI in Edge applications provide the technical ingredients to match performance, cost, and power to the computer vision design needs in the camera and edge gateways and servers.
- Acceleration: The right acceleration silicon can help ensure that highly accurate vision analytics are delivered efficiently. This doesn’t have to mean making costly investments in Graphics Processing Units (GPUs). Intel provides a comprehensive selection of options, including Intel® CPUs and Intel® Vision Accelerator Design Products based on Intel® Movidius™ Vision Processing Unit (VPU) and Intel® FPGAs.
- Software: Finally, the software provides the ‘wrapper’ that accelerates and integrates intelligent vision solutions and supports their delivery at scale. Intel® Distribution of OpenVINO™ toolkit is an end-to-end software suite that helps integrate vision across your customers’ infrastructure.
Having been swamped in buzzwords and marketing jargon about video analytics and AI for many years, SIs finally have some good news for their clients. The technology is catching up to their aspirations, and with the right approach to integrating better silicon architectures, accelerators and software, transformative outcomes are possible.
Read our latest blog to find out how computer vision supports the cities of the future
Or, check out our training video series: Integrating Vision into your Infrastructure
The post Getting Started with AI for Computer Vision appeared first on IoT@Intel.