December 2022 Release Notes
  • 23 Dec 2022
  • 3 Minutes to read
  • Dark
    Light
  • PDF

December 2022 Release Notes

  • Dark
    Light
  • PDF

Article Summary

A Note From Landing AI


Dear LandingLens community,

Thank you for continuing to collaborate with us on ways to improve LandingLens. Your continuous feedback and support have been central to this past year of development. Exciting things are being released over the next few weeks, and we can't wait to hear what you think! 

Thanks again for being part of the LandingLens community, and Happy Holidays!

From,
Landing AI

Announcement 

🚨 We temporarily removed the ability to delete Classes as we work on fixing a downstream issue it is causing with previously trained Models. We'll announce in the Release Notes when we turn this feature back on. We apologize for any inconvenience. 

What's New

  • If you've already trained your Model, LandingLens will now automatically make Predictions on newly uploaded images to that Model. This feature will save you time from needing to retrain your Model after you upload new images to your dataset.
  • The Models sidebar is now dynamic! Previously, when you updated your Ground Truth, you needed to retrain your Model to see new updates in the Models sidebar. Now, you can update your Ground Truth, and the Models sidebar will update automatically.
  • We now offer training Models with custom options so you can have granular control over pre-processing transforms and manual hyperparameter tuning.
    Available Augmentation Settings 
  • You can now save Models! You can then switch between Models to view their performance results. 
    Switch Models 
  • When labeling images in Object Detection Projects, you can start labeling immediately by drawing your first bounding box before you have any Classes created. After you've created your first bounding box, you can enter the name of the Class.
    You Can Draw Bounding Boxes Before Creating Classes 
  • We've updated the Project creation user interface (UI)! Note: This update is only applicable to Free Trial users.
    • There are now descriptions of each Project Type to help you pick the right one. 
      (Left to Right): Object Detection, Segmentation, and Classification 
    • You can start a Project quickly by loading one of our sample datasets. 
      Load Sample Data

Bug Fixes

  • Fixed: We squashed a bug that prevented images from displaying on the Historical Data page. [Internal Locator: CL-1336]
  • Fixed: We made sure that the colors of the Check Label Quality suggestions match the Ground Truth labels. [Internal Locator: LAN-8474]
  • Fixed: Similarly, the colors of your labels now match the Model's Prediction labels. [Internal Locator: LAN-8386]
  • Fixed: We fixed multiple bugs, so you can now upload images as expected. [Internal Locators: LAN-8484, LAN-8427, & LAN-8464]
  • Fixed: We changed the title of the "Defects" tab to "Classes". [Internal Locator: LAN-8458]
  • Fixed: Previously, if a device was already selected on the Live Prediction tab, users needed to select that device again when switching to the Historical Data tab. We fixed this so you can start working on your predicted images immediately. [Internal Locator: CL-1270]
  • Fixed: An error message no longer displays when you click the "Nothing to Label" button and go back to the Data Browser. [Internal Locator: LAN-8495]
  • Fixed: Images that have the "Nothing to Label" button applied will no longer display as "Unlabeled". [Internal Locator: USNAP-530]



Was this article helpful?