visage|SDK: Release notes

2023-11-24 visage|SDK 9.1 stable

Platforms: all

• Introduced new landmarks in lips, eyes, and eyebrow regions
  • In addition to additional landmarks, a small improvement in the accuracy of detected landmarks has been obtained.

• Significantly increased overall tracking precision by reducing jitter compared to 9.0
  • Additional reduction of jitter is available by enabling the image denoising feature. The feature has a significant impact on speed depending on the size of the input image; therefore, it is disabled by default but can be enabled in configuration settings.

• Improved overall 3D model fitting quality
  • General fitting quality has been improved with a small trade-off in fitting speed compared to 9.0. The system still works within real-time constraints.
  • Additional improvement in initial fitting quality for rotated faces has been obtained.

• Fixed VisageFeaturesDetector crashes when attempting to input an image containing multiple faces, while the number of detections is limited to 1

Platforms: HTML5

• Optimized memory usage
  • visage|SDK library now allocates half as much memory while still providing enough memory to run all visage|SDK modules in the same thread. This optimization prevents crashes in Google Chrome for Android phones with limited memory space for Google Chrome browser. operations.

• The system now works in strict mode

2023-08-02 visage|SDK 9.1beta4

Platforms: all

• Fixed degradation of alignment points accuracy in eye region for very high yaw rotations introduced in 9.1b2

2023-07-17 visage|SDK 9.1beta3

Platforms: all

• Significantly increased precision of face alignment by reducing landmark jitter by a minimum of around 75% compared to 9.1b2 and by a minimum of around 90% compared to 9.0
• Image denoising feature introduced to further reduce tracking jitter
  • The reduction in jitter is an additional 54% compared to when this feature is disabled in the current release.
  • The feature has a significant impact on speed depending on input image size.
  • The feature is optional and, as such, can be enabled or disabled via the configuration settings
• Improvement in speed of the whole face tracking pipeline due to optimization of the model fitting procedure
  • Improved compared to 9.1b2, but up to 2ms slower compared to 9.0, depending on the device. However, results are still within the expected frame rate for real-time performance.
• Fixed FaceDetector sample crashes when attempting to open an image containing multiple faces

2023-07-03 visage|SDK 9.1beta2

Platforms: all

• Increased tracking accuracy of lip region
  • Tracking algorithm tracks additional 16 points in the lip region
• Increased tracking overall precision
  • by reducing the jitter of landmarks by around 67%
• Improved fitting quality
  • by simplifying the flow of the algorithm and by increasing the range in which shape units are recalculated, with the trade-off of slowing down the execution speed of the track function by up to 4ms, depending on the device

Platform: HTML5

• Optimized memory usage
  • visage|SDK library now allocates half as much memory while still providing enough memory to run all visage|SDK modules in the same thread. This optimization is especially important to prevent code crashes in Google Chrome for Android on phones with limited memory space for Google Chrome browser operations.
• The system now works in strict mode

2023-04-28 visage|SDK 9.0.2 stable

Platform: HTML5

• Fixed errors in the FDP class
  • All static methods described in the FDP class documentation can now be successfully used
• Fixed library to work in strict mode

2022-12-21 visage|SDK 9.0 stable

Platforms: all

• New tracking model reduced in size
  • Retained accuracy of landmark points and gained higher precision (jittering)
  • Due to smaller models loading time decreased

Platform: HTML5

• Reduced library size
  • The size of HTML5 libraries has been reduced by ~25%, which allows for faster loading

2022-07-26 visage|SDK 9.0beta1.1

Platforms: all

• Optimize analyseStream() function results in case when only FaceAnalysis license is used

Platforms: HTML5

• Fix memory leak in VisageFaceAnalyser API

2022-07-15 visage|SDK 9.0beta1

Platforms: all

• New age and gender estimation models
  Introducing smaller, faster and more accurate age and gender estimation models.

• Multi-frame analysis module
  Introducing new API functions for multi-frame age and gender analysis to ensure more stable and accurate results

• Fixed multi-face tracking when the fitting is disabled

• New tracking and detection output parameter facial bounding box
  Introducing a new parameter for obtaining the bounding box of each tracked/detected face

• Slightly improved face tracking precision
  Improved physical contour tracking

Platforms: Android, iOS

• GPU support
  visage|SDK algorithms now can be offloaded to the GPU

Platforms: macOS

• Support of macOS ARM architecture
  visage|SDK package for macOS provides optimized libraries for ARM and x86_64 chip architectures

2022-02-10 visage|SDK 8.8.1

Platform: Android

• Lowered minimum supported API level to 21 (Android 5.0)

2021-12-14 visage|SDK 8.8 automotive edition

Platform: Linux

• Thoroughly tested stable version of visage|SDK for the in-cabin use case optimized for face tracking on near-infrared images

2021-12-14 visage|SDK 8.8

Platforms: all

• Liveness API exposure
  The usage of the Liveness preset actions is now also available through the FaceRecognition license.

• Upgrade of VNN algorithm for tracking and detecting masked faces
  Face tracking and detection algorithms are enhanced so that they can track and detect faces wearing protective masks of various colors and patterns.

• Removal of the legacy tracking and detection algorithm
  With the improvement of the quality and performance of the VNN algorithm, we have achieved state-of-the-art face tracking and detection. In order to simplify usage and reduce the data, all those algorithms that are no longer competitive are removed.

• Switching from visible to physical contour
  Introducing the physical contour the stability and accuracy of one of the main visage|SDK features – 3D head-pose estimation has been improved. Improved the usability of the visage|SDK for many market fields such as DMS, Virtual Try-on, and Gaming.

• Swift wrapper
  It is now possible to develop with the visage|SDK in Swift language on iOS and macOS using the newly implemented Swift wrapper

2021-09-24 visage|SDK 8.8beta2 automotive edition

Platform: Linux

• The initial release of visage|SDK for the in-cabin use case optimized for face tracking on near-infrared images.

2021-07-16 visage|SDK 8.8beta2

Platforms: all

• Face tracking and detection with protective masks. Face tracking and detection algorithms are enhanced so that they can track and detect faces wearing protective masks of various colors and patterns.

• Switching from tracking and detection of visible contour points to physical contour points. Introducing the physical contour the stability and accuracy of one of the main visage|SDK features – 3D head-pose estimation has been improved. Improved the usability of the visage|SDK for many market fields such as DMS, Virtual Try-on, and Gaming.

Platform: iOS

• Swift wrapper. You can now develop with visage|SDK using Swift language on iOS using a newly implemented Swift wrapper

2021-07-06 visage|SDK 8.7

Platform: Ambarella

• The initial release of visage|SDK for Ambarella. Offers single-face FaceTrack functionality.

2021-04-15 visage|SDK 8.7.1

Platforms: HTML5

• Licenses work as intended when the application ID is supplied.

Platform: Linux, RedHat

• Fixed possible inconsistencies with license validation.

2020-09-30 visage|SDK 8.7

• New face recognition model. Introducing a smaller, faster, and more accurate face recognition model.

• The new and improved face detection model. Introducing improved face detection model, more robust to various challenging conditions such as faces with high variability in scale, illumination, pose and occlusion.

• New fast mode of VNN tracking algorithm. Introducing new VNN algorithm fast mode which significantly improves performance at the cost of feature points precision while keeping the same precision of head pose.

• Improved tracking performance of VNN algorithm. VNN tracking algorithm now works with higher FPS, with significant improvements on devices such as high-end mobile and desktop devices.

• Reduced tracking noise in VNN tracking algorithm on mobile platforms – Android, iOS. New VNN tracking models now work with less tracking jitter.

2020-07-31 visage|SDK 8.7beta2

• The new improved face detection model. Introducing improved face detection model, more robust to various challenging conditions such as faces with high variability in scale, illumination, pose and occlusion.

• New fast mode of VNN tracking algorithm. Introducing new VNN algorithm fast mode which significantly improves performance at the cost of feature points precision while keeping the same precision of head pose.

• Improved tracking speed of VNN algorithm. VNN tracking algorithm now works with higher FPS, with significant improvements on devices such as high-end mobile and desktop devices.

• Reduced tracking noise in VNN tracking algorithm (iOS, Android). New VNN tracking models now work with less tracking jitter.

2020-05-08 visage|SDK 8.7beta1

• New runner on mobile platforms – Android and iOS. Introducing new runner – TensorFlow Lite, optimized for running neural networks and significantly improving the performance of visage|SDK algorithms.

• New Face recognition model. Introducing a smaller, faster, and more accurate face recognition model that completely replaces the old model that will no longer be distributed. The model is available for OpenVino™ runner on desktop platforms and Tensorflow Lite runner on mobile platforms.

2020-01-15 visage|SDK 8.6

• A novel experimental tracking algorithm – VNN – introduced

The new algorithm minimizes jitter, increases tracking accuracy and robustness but reduces tracking performance (speed). It is demonstrated in ShowcaseDemo and FaceTracker2 samples via Ultra tracking configuration.

• New neural network runner provided – Intel’s OpenVINO™

Significantly improves the performance of age estimation, face recognition and face tracking with VNN algorithm on Intel 64-bit processors.

• Ear tracking

Additional 24 feature points on ears are now tracked (12 points per ear). Ear tracking is configurable through the tracker configuration file or API.

• Iris tracking

Face data from tracker and detector now includes information about iris diameter.

• VisageConfiguration API introduced

It is now possible to modify specific tracker configuration parameters via an interface during tracking.

• Age estimation accuracy improved

2019-05-30 visage|SDK 8.6beta1

• Ear tracking introduced: Tracker can track 24 additional feature points on ears (12 points per ear); Configurable via configuration file.
• Iris tracking: Face data from tracker and detector now includes information about iris diameter.
• Age estimation accuracy improved
• Tracking performance (speed) improved

2019-03-20 visage|SDK 8.5

• Improved smoothing filter.

Smoothing of feature points is performed using multiple filters. For still face, higher amount of smoothing is applied while fast movements are less smoothed in order to avoid noticeable delay. Increased stability of feature points and head position especially in profile and half-profile pose.

• Refactoring of frame preprocessing resulting in more stable FPS and improved accuracy on high-resolution frames

The core tracking loop was reimplemented to make the tracking frame rate less dependent on the size of the face in the image. This fixes performance drops in cases where the face takes up a small portion of the frame. Additionally, noise introduced by resizing of higher-resolution frames is reduced which results in more stable tracking.

Platform: HTML

• API upgraded to use memory arrays. API for fetching tracking data has been modified to return memory arrays. Improves performance and simplifies memory management of tracked data.

Platforms: iOS, Android

• ShowcaseDemo introduces example of tracking from video including source code.