Release 1.20

Updated
  • Updated on Jul 24, 2025
  • Published on Jul 18, 2025
  • 2 minute(s) read
Prev Next

Release 1.20.216 (Bug fix release) - July 24, 2025

  1. Fix for an issue with providing feedback on data generation projects that existed before 1.20 release.

Release 1.20.213 (Bug fix release) - July 23, 2025

  1. Fix for an issue where anomaly detection sessions failed to run when the data was sourced from Inspection Studio(Vision Command) dataset.

Release 1.20.205 (Bug fix release) - July 18, 2025

  1. Fix for an issue with incorrect failure retries.

  2. Fix for issues with demo data setup.

Release 1.20.202 - July 17, 2025

Anomaly identification

  1. The anomaly identification models can optionally be provided with defect image samples with/without localization segment masks to enhance the result quality.

  2. Multimodal models that utilize text information support defect class text descriptions.

  3. Usability enhancements.

Model training

  1. Support for supervised training of a semantic segmentation model in addition to object detection and single-label classification models.

Catalog page enhancements

Directory-based browsing

Tabular mode enhancements

  1. Cleaner column names by grouping columns under individual tables in the view.

  2. A better presentation for column selection and an additional capability to reorder columns.

  3. Compact/Expanded mode toggle to optionally view columns like file_path fully inline in a multi-line display

Associate higher-level types with catalog columns

When importing a catalog in CSV format, provide a mapping from catalog columns to higher-level types (such as ground truth class, prediction class, etc.). This mapping will be used to enhance the catalog interactions in upcoming releases by providing type-aware behaviour. For example, when creating a view that requires a ground truth class column, the specific column mapped to a higher-level type will be selected by default.

  • For a new import catalog, the type mapping is presented as part of the import flow.

  • For existing tables, the type mapping can be added as part of the ‘Edit’ action under the ‘Table/View Actions’

  • For catalog import in standard formats, such as COCO, the type will be automatically associated with the columns defined according to the format standard.

Task type specific views

The Import catalog allows you to specify a task type (Classification, Object Detection, Segmentation) and content type (Ground Truth, Prediction), which enables the automatic addition of the imported table to a task-specific view or the creation of a new task-specific view. Overall, this flow minimizes the steps required to integrate your catalog into the dataset for a catalog that has metadata that maps to commonly used task types.

Vision Command enhancements

  1. A ‘compare’ mode has been added to compare images from two different points on the analytics results.

  2. Slideshow and other full-resolution page enhancements for Vision Command.

Other enhancements

Navigation panel enhancements

  • Navigation panel options have been rearranged for better grouping of related options.

  • Ability to pin frequently accessed options

Add custom tags

In a visualized job, select and add custom tags on images. Use these tags to filter and group images, and download these tags as part of the result set download.

  • Add custom tags on the thumbnail or through bulk selection.

  • Use custom tags as grouping criteria to colour points on the plot view.

  • Use custom tags to set filter conditions.

  • Download these tags as part of the result set download.

Full resolution view enhancements

A refreshed user interface to work with a full-resolution view.