Forward Vision:
How TensorIoT is using Amazon Lookout for Vision to improve our Solutions
By: Nicholas Burden, Technical Evangelist, TensorIoT and Viraj Deshwal, Machine Learning Engineer, TensorIoT
At re:Invent 2020, Amazon Web Services (AWS) debuted Amazon Lookout for Vision, a new machine learning (ML) service that provides industrial customers with an easy and cost-effective method for detecting visual defects on production units and equipment. As official launch partners for Amazon Lookout for Vision, TensorIoT is excited to share one of the ways we’re using this innovative technology to improve our solutions. In this blog, we’ll take a closer look at the tools Amazon Lookout for Vision provides and how we are incorporating them into our SmartInsights industrial analytics solution.
How does Computer Vision Machine Learning help Manufacturing?
Using computer vision (CV) to detect defects offers multiple advantages. CV can provide immediate feedback for removal or remediation of products during the manufacturing. CV can also be incorporated into existing human quality assurance (QA) processes at different levels, such as a first line for obvious defects or process errors, second line verification of human-detected errors, and acting as an objective assessment tool to standardize areas prone to natural human variance. By integrating CV with your other systems, you unlock additional uses and added value. For example, integrating CV into ERP Systems gives a more accurate view of viable parts availability based on your live throughput, defect rate, and current and forecasted order demand.
What does Amazon Lookout for Vision do better?
Catching defects in the production process is critical for any successful industrial operation. Human inspection can become unreliable due to operator fatigue or distraction, making computer vision attractive as an adjunct to human visual inspections in a production environment. Prior to the release of Lookout for Vision, computer vision machine learning often relied on coded rules and imposed strict restrictions on camera positioning. Lookout for Vision helps to automate visual inspection of product defects across production processes using deep learning models, and Lookout for Vision adjusts to differences in camera angle, lighting and other complications in the operational environment better than prior methods. By using Lookout for Vision to reduce the need for carefully controlled environments, computer vision solutions can be employed in more diverse operations.
Integration with SmartInsights
SmartInsights is an AWS-born IoT platform that allows operators, managers, and executives the ability to quickly gain actionable insights into the status of their industrial operations, pulling data directly from machines on-site. SmartInsights delivers improved situational awareness of industrial operations, more meaningful insights for workers and managers to quickly undertake corrective action, and more confident decision making both in day-to-day operations and production planning.
With Lookout for Vision in place, not only will QA specialists and other production workers have an immediate interface to interact with to see results and make decisions, they’ll also be able to correlate the detected defects with concurrent machine data to identify patterns and fine tune machine recipes to optimize between throughput and defect rate to maximize defect-free output. By integrating Amazon Lookout for Vision with TensorIoT’s SmartInsights solution, manufacturers and other industrial customers gain increased operational efficiency by using additional sources of data for visualization and analysis. For example, anomaly detection from Lookout for Vision could be correlated with variations in other production equipment data captured by SmartInsights and used in tandem for improved predictive maintenance scheduling.
Moving forward, TensorIoT plans to continue incorporating the latest AWS technology into our solutions and keep making things smarter. To learn more about SmartInsights, contact us at demos@tensoriot.com.
