Product Overview: How does our Visual Quality Inspection work?
By: Dr. Audrey Chung, Jake Walker
How Quantitative Explainable AI can make human visual inspections more productive
Quality inspection in manufacturing facilities is an imperative process but can be tedious for visual inspections. Since repeated visual inspections can be mundane, this can lead to missed defects. Humans are experts, but aren’t perfect. A human inspecting the same product part hundreds of times a day can lead to fatigue, which can lead to unintended complacency, which can lead to an imperfect inspection process. Complex product inspections can take time, sometimes hours. Therefore, all of these factors make the human inspection process an imperfect process.
We believe using AI-assisted visual quality inspection makes the process better for the inspector. AI will not experience fatigue, and can also process images of highly complex inspections in seconds, if not milliseconds.
DarwinAI’s platform aims to support that inspection process using neural networks that emulate the cognitive capabilities of the human brain. As more manufacturers are considering AI, the question we get most often is how does it work?
DarwinAI provides an end-to-end deep learning-based visual quality inspection solution for manufacturers to improve product quality and reduce production costs by easily integrating AI into their production process. Using images of your parts and our proprietary Explainable AI technology, we specialize in building accurate, efficient, and robust AI with less data than conventional AI methods to help manufacturers realize faster time to value from making the switch to using AI. Whether you want to automate a visual quality inspection checkpoint so you can reassign your inspectors to more complicated tasks, or you want to bring in some AI assistance to streamline the work your inspectors are doing to ramp up to scale — we can help.
Our product, DarwinAI Visual Quality Inspection, comprises three main parts: an image capture solution; our DarwinAI Inspection Platform; and an inspector user interface.
Image Capture Solution
Every good deep learning AI starts with good data, and our product is no exception. The first part of DarwinAI’s Visual Quality Inspection is an image capture solution that works for a manufacturer’s specific visual inspection problem. DarwinAI collaborates with manufacturers to create an image capture solution that is customized for the manufacturer’s specific needs; generally speaking, an image capture solution consists of one or more cameras, lighting recommendations as needed for the production floor, and a rig to hold and move the part during imaging. Conversely, DarwinAI can work with manufacturers to integrate an existing imaging solution, or work with a preferred system integrator for visual inspection processes that require more complicated automatic part handling.
DarwinAI Inspection Platform
Data collected through the image capture solution is passed to our DarwinAI Inspection Platform to be analyzed by a custom-built deep learning AI model. The main components of the DarwinAI Inspection Platform enable manufacturers to find defects using AI, present insights behind each AI decision, ensure defect traceability by storing and retrieving the captured image data and corresponding defect results, and continuous AI improvement by collecting feedback from the QA inspector to boost future AI performance. Lastly, the DarwinAI Inspection Platform includes an inspector interface for easy and intuitive interaction with our solution. These components are connected by a customized external device controller, which handles the communication and transfer of data between the image capture solution, the DarwinAI Inspection Platform, and the inspector user interface.
Inspector User Interface (UI)
The inspector interface is the final part. Through this user interface, an inspector can start and stop the visual quality inspection solution, view the captured image data along with the AI defect decision(s) and relevant insights, and provide expert feedback (e.g., whether the AI defect decision was correct) to boost the performance of the deep learning AI in the future. The inspection results are communicated back to the Inspector via the Inspector UI in real-time (if there is someone actively monitoring the inspection process) or as a summary (if the inspection process is fully automated). The inspector interface also provides a summary overview of the visual quality inspection solution, allowing manufacturers to better understand DarwinAI Quality Inspection’s impact on manufacturing productivity and process efficiency.
During the visual inspection process, a new part arriving for inspection will trigger the image capture solution; this image capture can be started manually if an inspector is required to monitor the process, or can be triggered automatically. The images are transferred to the DarwinAI Inspection Platform, where custom-built AI(s) find and localize any defects and store the results for process traceability. Every quality inspection solution is going to be different: different ways of part handling, different lighting, and different environment characteristics. DarwinAI has a team dedicated to overcoming these challenges so that the full benefits of our unique deep learning platform can be realized. Maximize your ability to find defects in complex inspections to ensure nothing defective gets into your customer’s hands. Improve your productivity. Deep learning can be implemented to be either entirely autonomous or support your inspectors by taking seconds to inspect a part that might normally take hours. Scale our platform within your organization to maximize value and proliferate ROI.
To learn more about how DarwinAI can help your manufacturing process, contact us or download our whitepaper “How AI delivers business value for today’s manufacturers.”