Face AR SDK v0.34.1: Improved Skin Smoothing & Effects Editing
Our Face AR SDK release v.034.1 features updated neural networks for iOS devices that can now perform skin smoothing of the person's forehead with greater accuracy. Moreover, effect designers can take advantage of extra processing options in Banuba Viewer to apply multiple Face AR effects to a group of images at the same time. Read our blog post to learn more about these and other changes.
Be the first to test new features of Face AR SDK!
Improved neural networks for skin smoothing on iOS devices
Skin smoothing is an important feature of many photo editing and beautification apps. Their touch up functionality will allow users to make great selfies by improving the look of their face in the best possible way. The technology can retouch the skin tone to make it look healthier and younger.
In the latest release of our SDK, we've improved the functionality of neural networks which perform skin smoothing on iOS and MacOS devices. Now the person's forehead looks more natural and smooth. This update resulted in the improvement of the segmentation quality and the decreased number of false positives and segmentation errors.
Skin smoothing based on neural networks works on any skin color and is compatible with other Face AR features like virtual makeup or face filters.
At the moment, these neural networks are only suitable for photo editing purposes. The real-time version will be available later this year.
How our neural networks perform skin smoothing
Skin smoothing. Example 1
Skin smoothing. Example 2
Extra options for processing AR effects in Banuba Viewer
Designers can take advantage of extra processing options in Banuba Viewer and apply them to a group of images with just a few clicks of the mouse, saving both time and effort. This feature will especially be useful for applications containing a larger number of constantly updating Face AR effects.
Added Java 8+ API desugaring support for Android;
Added New Action Units effect with background segmentation
Added static tensorflow lite version for Android;
Enabled RGB camera on devices with Snapdragon 625 which will improve performance on this chipset;
Changed tflite_runner different delegates support for each feature;
Changed processed images location in Banuba Viewer.
Changed the distribution of EffectPlayer for iOS as xcframework