What's new

The following functions, features, and support have been added for PowerAI Vision Version 1.1.4:

Action identification
You can train models to identify actions in a video. To get started, see Labeling actions.
PowerAI Vision Non-production edition update
This edition has updated functionality. The number of deployed and trained models is limited, but you can now export models. See PowerAI Vision Non-production edition for details.
Track usage metrics
Administrators can track how many inferences have been run in PowerAI Vision. For details, see Monitoring usage metrics.
PowerAI Vision trial version
The functionality of the trial version has been updated. The number of deployed and trained models is limited, but you can now export models. See IBM PowerAI Vision Trial for details.
Inferencing for videos with Detectron models
Detectron models now accept videos for inferencing; not just images.
View video inferencing results for Detectron models
After a video inferencing completes, the annotated video plays. For details, see Testing a model.
Updates to the user interface
The following changes have been made to the user interface to improve your experience:
  • The user name drop down menu in the navigation bar now contains the following:
    • The log out action
    • Edition information (time trial and Non-production versions only)
    • Product version
  • A refresh button has been added to the GPU usage information.
  • If you have external GPUs, they are now included in the GPU usage list. External GPUs are those that are used for processes outside of PowerAI Vision. For IBM Cloud Private installations, all GPUs are listed as External.
  • When auto capturing frames, you can now specify values as small as .001.
  • Notifications will be sent that show progress on active trainings and video inferences.
  • Columns for Original file name, Actions, and Augment were added to the Data Set Details grid. The original file name will not be accurate, or will be missing for older data sets that are imported. Additionally, the original file name is not available for captured video frames or augmented files.
  • An "Auto labeled" filter has been added to the Data Set Details side bar, which can be used to hide or show images or video frames that have auto labels.