What's new in Maximo Visual Inspection 8.9

Group images

The Group images feature uses efficient data pattern insights to help you explore and understand the images in your large data sets.

When you select Group images from the menu bar, the feature vectors in your data set images are extracted. Data summarization techniques use the extracted feature vectors to identify possible patterns in your images. The patterns that are identified represent the images that contain similar features. These images are collected into the number of image groups that you selected.

You can then review the image groups to determine whether your images are diverse enough to train a model. You can also move or copy the image groups into existing data sets and use them to retrain and improve the accuracy of a model.

Metrics for data pattern insights on the Grafana dashboard

A row for data pattern insights on the Grafana dashboard includes the following metrics:

• Data set feature extraction histogram
• Data set summarization histogram
• Mean data set feature extraction duration
• Data set feature extraction count
• Mean data set summarization duration
• Data set summarization count

Workload scale customization

To help improve efficiency and reduce costs, you can scale Maximo Visual Inspection to use more or less resources. You can adapt scaling according to the size of your application setup and its corresponding workload. Scalable resources include memory, CPU, and replicas.

For more information, see Workload scale customization.

Object storage

Object storage is a persistent storage system that organizes data as distinct units, called objects. You can enable object storage for use on inspection data sets. Object storage provides the following benefits for your inspection data sets:

Fast data access

Each object contains data, metadata, and an identifier that helps you to quickly access and retrieve data.

Virtually unlimited storage space

Keep adding data to an object storage system without limit.

Simple data organization

Objects are saved in a flat data environment, where you no longer need to organize data in complicated hierarchical structures, like folders and directories.

For more information, see Object storage.

Deployment target options for anomaly model training

When you are training an anomaly model, you can select a deployment target that is suitable to your use case and deployment device. The deployment target options include the following device types:

Server

The most accurate but uses the most memory and runs the slowest.

Edge

Uses the least memory and runs the fastest. It can lead to reduced accuracy.

Hybrid

The middle option between the server and edge device types.

Automatic dimensionality reduction

Dimensionality reduction can reduce the dimensions of feature embeddings. This option was introduced in Maximo Visual Inspection version 8.8. It is now automatically applied at deployment time in accordance with the deployment target that was selected at training time.