AI-assisted computer vision is key to anomaly detection, a process used to automate quality inspections on production lines, help medical professionals diagnose conditions, and monitor retail inventory and store traffic. When it comes to manufacturing, computer vision also can help identify defects and keep on-site workers safe.
To increase the speed and simplicity of creating these computer vision solutions, Intel and Microsoft partnered with Scalers AI have released new factory solution accelerators available on GitHub. With this sample code, developers and solution builders can quickly set up a manufacturing anomaly detection system and a worker safety solution. Additionally, they serve as templates for other computer vision uses as well.
Using computer vision for defect detection and worker safety
The anomaly detection solution is designed to identify manufacturing flaws in pump impellers, quickly rejecting those that are outside of acceptable tolerance ranges. This solution accelerator also can be trained to inspect a wide range of other products. This use of computer vision is a common one, as human eyes usually take longer to find flaws—some of which are important but easy to miss while performing this type of repetitive, exacting inspection. Even if a missed anomaly doesn’t cause a product to malfunction, it may cause a product such as a pump to perform poorly and potentially require repair over time. Additionally, using computer vision for inspections allows employees to perform more value-added work in a manufacturing setting.
The AI-based worker safety solution available on GitHub can monitor potentially dangerous areas of industrial settings and shut down or suspend the operation of nearby machinery if a person is detected in that area. Again, this model can be modified for a variety of safety purposes and solves a challenge that would otherwise require constant human monitoring.
These solution accelerator sample codes are designed for optimal operation on Intel-powered computers with iGPU acceleration and running Microsoft Azure services. A series of training videos will guide you through various steps for setting up and deploying these accelerators.
Leveraging EFLOW and OpenVINO to optimize development
The Scalers AI solution accelerators show how you can build and train AI models in Microsoft Azure on compute devices with Intel processors deployed at the edge. The accelerators also integrate the Intel® Distribution of OpenVino™ toolkit with Azure IoT Edge for Linux on Windows (EFLOW). This simplifies the creation and registry of machine learning containers and device provisioning.
By running Linux IoT Edge modules on Windows devices, you can run Windows applications and Linux modules on a single device, reducing the overhead and cost of separate devices for different applications. EFLOW doesn't require extensive Linux knowledge and utilizes familiar Windows tools to manage your EFLOW device and workloads. In addition, Windows IoT provides trusted enterprise-grade security with established IT admin infrastructure. Lastly, the entire solution is maintained and kept up to date by Microsoft.
The Intel® Distribution of OpenVINO™, meanwhile, allows developers to optimize, tune, and run comprehensive AI inferencing using model optimizer and runtime development tools. This toolkit eases the deployment of computer vision and deep-learning models and applications.
By leveraging EFLOW and the Intel® Distribution of OpenVINO™ and these solution accelerators together, you can cut countless hours in learning how to deploy solutions that include streaming video analytics, AI, times series data, and more. The solutions are designed to deploy on either new or existing hardware and infrastructure as long as it's running the relevant Azure IoT platform and other tools listed above.
Explore more Intel and Microsoft computer vision technology
The first step to kicking off development of these computer vision solutions is to download these smart factory solution accelerators and get started on setting them up for your particular use. A fast, introductory augmented reality (AR) experience that leads to the accelerators also is available. To dig deeper into the expanding real-world applications across a range of industries for Intel and Microsoft-enabled computer vision solutions, download this whitepaper.