January 18, 2024
Last updated
Last updated
Improvements in the model-sharing experience and Deepomatic models page
The search feature has been made more typo-tolerant. In particular, case sensitivity and accent sensitivity have been removed. In the mobile app settings, it is possible to configure whether the search is typo-tolerant or not.
In the mobile app settings, field workers can now choose whether they want to display or not work orders 100% completed. By default, all work orders are displayed.
The progress_score is now updated when the work_order_types key is patched. Once the work_order_types of a work order are modified, the progress_score is therefore computed taking into account the tasks included in the new work_order_types set.
It is possible to use URLs in order to analyze PDF files. Now, both Base64 and URLs are accepted to analyze PDF files.
Several Export <> Import issues are fixed:
View conditions can flawlessly get re-imported
The val/train split of images is respected in the import
Renaming a view in the original project will not create any issues in reimporting the project
Several count-related issues are resolved and a few improvements are made:
The image distribution count on the overview piechart is corrected and improved to include the counts of total and unannotated images in val/train sets.
The per-concept distribution chart accurately displays the number of annotations, and detection views it allows you to toggle between annotation and image counts.
The default data augmentations in training are removed to give you the possibility to choose the required argumentation
In the connectors (Camera and Work order), the callback URL was not working after the page was refreshed. The issue has been fixed.
