Bring AI into your operations to detect anomalies

• Identify the use case

  IBM® Maximo® makes computer vision a reality for users without coding expertise.

  To get started, it’s important to understand the business case you want to use visual inspection for. This could be quality control, defect detection, asset and infrastructure inspection, or any other application that involves visual analysis.

  Then, define the purpose and objectives of the visual inspection process. Identify the types of assets or objects to be inspected, the specific defects or anomalies to be detected, and the level of accuracy required.

• Capture images

  Building the computer vision model starts with images. These images can be taken from inspections completed offsite and then assessed collectively. Cameras and other edge devices can automatically monitor your assets more frequently and capture key images you need to build the AI computer vision models. In addition, the cameras may be used to detect issues in real time, using a deployed model option.

  It’s important to remember that the quality and quantity of training data can improve the accuracy and performance of your models.

• Quickly label the data

  The user can quickly label the data with easy-to-use visual tools to define a wide variety of both good and bad conditions.

  Be sure that the data adequately covers different variations, conditions and anomalies.

• Autolabel and image augmentation

  Once a few images are successfully labeled, IBM Maximo can autolabel a much larger data set, saving a tremendous amount of time. Also, by using image augmentation, you can create new images from your existing data set by adding noise, blur, rotation and other effects to improve the accuracy of the AI models.

• Build and train the AI visual model

  There are many predefined models to choose from, depending on your use case. For example, there are models for image classification, object detection, image segmentation and action detection that are ready to use right out of the box.

• Check the accuracy

  Maximo provides built-in performance metrics that help quantify the model's accuracy. These metrics include measures such as accuracy, precision, recall and F1 score.

  You can develop highly accurate visual models without the need for data scientist skill sets. Even so, data scientists will also appreciate the ease of use and simple access to many advanced optimization tools.

• Deploy the model at the edge

  Computer vision models can be developed and deployed in a few hours—not weeks or months. Models can also be deployed at the edge with the click of a button. This feature lets organizations constantly detect issues and take immediate corrective action so small issues don’t become major outages.

  By offering flexible integration options, Maximo lets you connect cameras, work order systems, IoT devices and quality management systems to create a comprehensive end-to-end solution for visual inspection.