Release 1.15

Release 1.15.100 (Maintenance Release) - December 8, 2023

1. Bug fix for a job resubmit failure on an autolabelling job.

2. Bug fix for an issue where cloning an autolabelling job resulted in the parent job’s auto labels being overwritten.

Release 1.15.92 (Maintenance release) - November 28, 2023

1. Improvements in autolabelling accuracy for classification jobs.

2. Bug fix for the issue where the similarity search and text search were not giving correct results for job sizes in the 25000-50000 point range.

3. Performance improvements in similarity search and text search.

4. Support for autolabeling jobs in a dataset created through the 'Import public dataset' operation.

Release 1.15.69 (Maintenance release) - November 14, 2023

1. Bugfix for an issue with autolabelling not labelling all the points in a job.

Release 1.15.67 - November 11, 2023

1. Autolabeling: Label your image data through machine-generated labels at 10x lower cost and up to 40x faster than manual labelling. Please get in touch with us at support@akridata.ai for more details.

2. UI enhancements:

   1. Filtering controls for ground truth and prediction classes: Filter the ground truth and prediction class values whose bounding boxes will be shown on thumbnails and the full-resolution views. This helps focus analysis on a few classes at a time.
      

   2. Fixed bounding box color for each class:  Within a visualization job, the bounding box color for a given class is fixed across all images.

   3. Class-wise precision and recall values: In an analyze job, the confusion matrix row and column headers are enhanced to show each class's precision and recall values. This gives a consolidated view of precision and recall values across all classes in a single panel.
      

   4. Sorting by prediction confidence: In an analyze job, if thumbnails are viewed in the statistics panel, then a control to sort the thumbnails by prediction confidence is provided.