March 2022
Interview

AutomatedBuildings.com

Babel Buster Network Gateways: Big Features. Small Price.
Control Solutions, Inc. - Minnesota

(Click Message to Learn More)


Transitioning from computer vision to DNNs

Using Deep Neural Networks (DNNs) for traditional computer vision (CV) workloads has several advantages.
  Ken Sinclair Founder, Owner, Publisher AutomatedBuildings.com

EMAIL INTERVIEW

David Kuo is the Senior Director of Product Marketing and Business Development at Mythic Inc, a manufacturer of high-performance and low-power AI accelerator solutions for edge AI applications. He is working to bring Mythic’s innovative analog compute technology to the industrial machine vision, automation, and robotics markets. David has over 20 years of experience in product marketing, product management, and business development experience in Industrial/Consumer IoT, Mobile, and Consumer Electronics markets.”

`

                    



Articles
Interviews
Releases
New Products
Reviews
Securing Buildings News
Editorial
Events
Sponsors
Site Search
Newsletters
ABB
Archives
Past Issues
Home
Editors
eDucation
Secured by Cimetrics
Links
Software
Control Solutions, Inc

Control Solutions, Inc

  1. Sinclair Can you talk about how the industry is transitioning from computer vision to DNNs and what the advantages of DNNs are?

davidDavid Kuo Using Deep Neural Networks (DNNs) for traditional computer vision (CV) workloads has several advantages. Implementing CV requires a unique vision and programming expertise that is oftentimes difficult to find. The algorithms developed are specific and not easily transferrable to other CV workloads, are not resilient to changes in the operating environment (lighting conditions, for example), and require significant development time before deployment. For example, to find a car in an image, the CV algorithm will look for the features of the car - the wheels, the doors, etc. - and then this effort is repeated for anything that may be captured in the same vicinity of a car. These could include a motorcycle, a bus, people walking the street, and more. The algorithm's accuracy also will be severely impacted if the image is dark or slightly out of focus. Depending on the application, the development time may take up to six months to ensure each potential image is captured and accounted for in the CV algorithm. Conversely, one DNN can be taught to detect many objects by training it with thousands of images (dataset) representing the targeted objects. The more relevant images used for training, the more resilient and accurate the DNN will become for the application. DNNs are more accurate with higher performance when compared to traditional computer vision and a system engineer can train and deploy a DNN model in weeks as opposed to months of CV development.

 

  1. Sinclair What has driven these advancements in AI processing over the last few years?

David Kuo There has been significant research in DNN algorithms for vision analytic applications in the past decade. In addition, advancement in inference processing hardware solutions makes it feasible to deploy these DNNs in edge devices. An example is an innovative technology such as analog compute-in-memory that provides high-performance DNN inference processing at low power envelops, making it possible to deploy complex DNNs to edge devices.

 

  1. Sinclair In industrial settings, how are AI-powered robots and autonomous forklifts increasing productivity? 

David Kuo With the growth of eCommerce and worker shortage caused by the COVID-19 pandemic, AI-powered robots are an essential part of a company's automation strategy to improve productivity and reduce production costs. Autonomous mobile robots (AMRs) can be found fetching products from storage racks and delivering them to human workers for processing in a distribution center, for example. Additionally, an autonomous forklift will deliver stacked pallets from the end of the assembly line to the loading dock in a warehouse. These robots are efficient, always on-time, and can operate 24/7 making them vehicles for constant productivity.

 

  1. Sinclair What about security systems and drones using computer vision in commercial and industrial buildings?

David Kuo With AI and CV, the security system will provide more specific information for operators to take action. Instead of notifying the control room that people have been detected, a security camera can run a trained DNN to identify if each worker is wearing the proper protective gear or if a worker has fallen on the production floor. Drones running DNNs can also be deployed to identify malfunctioning equipment in remote locations or to environments not accessible for human operators.

 

  1. Sinclair What does the future hold for AI applications in buildings, and what can we expect in the next three-to-five years?

David Kuo As can be seen responses above, AI applications will play a significant role in building operations where AI is deployed as part of an ongoing automation strategy to improve productivity, drones and robots are part of a predictive maintenance system, and AI-powered security systems provide more specific data to improve worker safety and facility security. This is part of the Industry 4.0 revolution coming to industries worldwide.

 

 

 





footer

opsys
[Click Banner To Learn More]

[Home Page]  [The Automator]  [About]  [Subscribe ]  [Contact Us]

Events

Want Ads

Our Sponsors

Resources