Release 1.16

Release 1.16.201 (Bugfix release) - April 02,  2024

1. Fix for Labeling panel showing a repeat of the same image if an image has bounding boxes from more than one stats panel cell.
2. Fix for default explore job creation failure in some cases for the first dataset created after a new organization registration.
3. Fix for issue with automatic type inference for boolean column type in import catalog operation.
4. Fix for the issue where the default view(all_pipeline_tables) update fails when a new pipeline is attached to the dataset with an imported catalog table.
5. Restrict the maximum number of login sessions per user to 8.
6. Update copyright year to 2024.

Release 1.16.197 (Bugfix release) - March 18,  2024

1. Fix for confidence score appearing wrongly on a 3D ground truth bounding box.

Release 1.16.196 (Bugfix release) - March 7,  2024

The release contains the following bug fixes

1. In some corner cases where tunables/filters are applied to reduce search space, the plot view for search results was presented incorrectly.
2. Incorrect 'default explore job not present' message in certain cases.
3. Increase the font size of the confidence score legend on thumbnails.
4. A corner case issue where asynchronous operations like submitting a job may return successfully without triggering the necessary processing.
5. Bounding boxes do not appear correctly on full-resolution image view when the 'context filter' is toggled off.

Release 1.16.179 (Maintenance release) - February 19,  2024

1. Autolabeling
   1. Autolabel accuracy improvements.
   2. UI enhancements
      1. Show bounding box on image sample thumbnails in labeling specification.
         
         
      2. An improved full-resolution image view when uploading image samples in labeling specification with a larger image area.
         
      3. Performance improvements when interacting with a labeling job with 100K+ annotations
   3. Performance improvements in creating resultset out of a labeling job with 100K+ annotations.
   4. In a labeling job, add approved/rejected/pending review state annotations to an existing resultset in addition to creating a new resultset.
      
   5. Export a labeling spec and create a new spec from this exported file.
   6. In a labeling job, when full resolution view is from a labeling tab, then the bounding boxes shown on full resolution image are filtered based on the selected row, column or cell in the labeling tab. This filtering can be turned off using the context filter toggle marked below.
      
2. Patch search performance improvements.
3. Enhancements in importing COCO format files to the catalog.
   1. 'license' key is made optional.
   2. The COCO bounding box format has a new virtual column type named AKD_BoundingBox2DCOCOJSON to use in Data Explorer jobs directly.
4. In an analyze job, view F1 score metric for each class on the Precision-Recall panel.
   
   

Release 1.16.139 - January 29,  2024

1. Autolabeling: Label your data automatically using the combined power of best-in-class LLMs(large language models) and computer vision foundation models.
   1. Define your labeling specification using the power of LLMs.
      
   2. Add image prompts with bounding boxes to enrich your labeling specification.
      
      
   3. Select images to label from catalog and fire off a labeling job.
      
   4. Interactively review the results and approve labels.
      
2. Simplified dataset registration: To get you up and running with just a few clicks
   1. Fill in data container details inline without losing your flow.
   2. Pre-selected data processing pipelines for most common cases.
   3. Trigger pipeline runs using Data Explorer compute infrastructure.
   4. Have your first visualization job automatically created and ready to explore.
      
3. Additional dataset actions: One-click actions for
   1. Executing all attached pipelines.
   2. Creation of a default explore job on all points of the dataset(within plan limits)
   3. Viewing the default explore job
      
4. Import catalog enhancements
   1. Import your catalog CSV file with auto inference of column types and table creation.
   2. Native support for input as a COCO format JSON file
      
5. Search enhancements
   1. Create a new search by combining multiple saved searches
      
   2. Additional filter to search within a resultset or exclude resultset points.
      
6. UI enhancements
   1. Full-resolution image page enhancements - Show the most relevant fields and move other fields to a detailed view
      
      
   2. Fine-grained reporting of pipeline progress.
      
   3. Confidence-based filtering of bounding boxes in analyze and auto-label object detection jobs.
      
7. Number of classes supported in Analyze jobs increased from 25 to 50.