Released 30 March 2022
FaceSDK 8.0 Release Notes
This release introduces anti-spoofing features and more:
- Passive liveness detection
- Samples for active liveness detection
- Web camera support for Linux (with v4l2-compatible web cameras)
- x86_64 build for Android
- LiveRecognition sample for Linux in C++/GTK+
- Improved detection of small faces
- Updated .NET and Java wrappers for better compatibility with different .NET and Java versions
- Updated Python wrapper for better compatibility with IDEs
FaceSDK 7.2.1 Release Notes
This release introduces a Mask-on Face Detection and the support of Apple Silicon processors:
- Mask-on face detection model added
- Support of Apple Silicon processors (arm64) in MacOS added
- TrackerMemoryTool.py utility's bugs fixed
FaceSDK 7.2 Release Notes
This release introduces thermal face detection and contains significant improvements:
- Thermal face detection is added. Use it with thermal cameras to detect faces to check their temperature (for fever screening) or to find out if a person is alive (for anti-spoofing).
- Face detection models could be loaded from a file (for example, thermal face detection model).
- Face enrollment speed has been increased 2X on Linux.
- Auto rotation of loaded JPEG photos using EXIF is added.
- Support for higher InternalResizeWidth values is added to detect even smaller faces on larger images and high-resolution streams.
- Configuration of face trimming is added to determine which faces should be excluded from detection.
- Full Python support (wrapper and samples) is added.
FaceSDK 7.1 Release Notes
This release is about major performance improvements and optimizations.
- Major improvements to face detection speed: up to 6x improvement on Android devices, up to 30% improvement on desktop computers.
- Optimized facial feature detection on all platforms, up to 2x faster on iOS.
- Optimized face enrollment on all platforms, up to 3x faster on Android.
- Improved compatibility with Java and bug fixes for Java samples.
- iOS development: a Swift sample added.
FaceSDK 7.0 Release Notes
- Revamped face detection engine with improved detection rates. Significantly improved detection of faces in complex lighting conditions, blurry and noisy streams, as well as angled faces.
- Optimized face matching with FSDK_MatchFaces on Android with nearly double the speed.
- Improved false acceptance rate when comparing angled and blurry faces.
- You can now target 64-bit versions of Android thanks to native arm64 support.
- Improved Xcode 10 compatibility.
See the Migration chapter in the documentation for details.
FaceSDK 6.5.1 Release Notes
FaceSDK 6.5.1 increases the speed of enrolment by 20% on Windows, macOS and Android, and by 5% on iOS and Linux. It also improves the accuracy of face recognition under very low lighting conditions. The 6.5 version might produce higher false acceptance rates on photos taken under very low lighting conditions. If you were using this version with regular photos, you likely did not notice. However, we recommend that you update to 6.5.1, which gives accurate results even with very low-lit photos. See the Migration chapter in our documentation for further details.
FaceSDK 6.5 Release Notes
1. FaceSDK 6.5 includes a new and improved face recognition engine, offering up to 50 times the matching performance and reaching unprecedented precision.
- Improved recognition rate of 99.85% according to NIST FRGC testing (up from 93.29% in the previous version) at false acceptance rate of 0.1%.
- Recognition engine is much more robust to different lighting conditions, as well as tilted and turned heads.
- Face search in a database is much more precise.
- Face matching speed in parallel threads is increased up to 20 times on Windows, up to 25 times on macOS and iOS, up to 50 times on Linux and up to 3 times on Android.
- Face template size is now 1040 bytes (down from 13324 bytes in the previous version).
2. Face detection speed is 20% higher on Windows.
3. FaceSDK 6.5 adds support for ARM-based Linux systems, allowing the use of FaceSDK-based apps in embedded Linux platforms including Raspberry PI 2+.
To achieve the accuracy increase, it was necessary to change the format of the face template. See the Migration chapter in the documentation for details.
FaceSDK 6.4 Release Notes
- New face detection engine works 2.5x faster in iOS/Android/Linux apps, 5x faster in Windows and Mac apps.
- VB6 camera stability improvements.
- The installer includes Visual Studio 2017 x86 and x86_64 runtime.
FaceSDK 6.3.1 Release Notes
FaceSDK 6.3.1 is a small update for version 6.3 that increases the accuracy of facial feature detection on Windows and Mac systems when using certain face detection parameters. If you’ve used FaceSDK 6.3 in your Windows or Mac apps, it’s recommended that you update.
What was improved? If you were setting HandleArbitraryRotations or DetermineFaceRotationAngle parameters to true on Windows or Mac (whether using the FSDK_SetFaceDetectionParameters or changing Tracker parameters), facial features of faces that were considerably rotated in-plane could have been detected with less accuracy. This may have lowered recognition rates and decreased the accuracy of age, gender and facial expression detection. If your photos were mostly upright, you likely did not notice this. Our live recognition samples that work with the camera set these parameters to false, so their behavior is not changed. However, if you work with faces using the default detection parameters (as in FacialFeatures, Lookalikes and Portrait samples), you may notice the accuracy increase.
FaceSDK 6.3 Release Notes
FaceSDK 6.3 is about low-level optimizations and performance improvements, as well as age recognition.
- Added age recognition functionality (the new "Age" attribute is now supported by the FSDK_DetectFacialAttributeUsingFeatures and FSDK_GetTrackerFacialAttribute functions).
- Added sample code for age recognition (the new "AgeGenderRecognition" sample).
- Major performance improvements through low-level optimizations.
- Facial Feature Detection time improvements (not including face detection stage):
- Windows: 10 times faster
- macOS: 4.5 times faster
- Linux: 3 times faster
- Android: 2.5 times faster
- iOS: 2.5 to 3.5 times faster
- Facial feature detection optimizations help further reduce jitter of facial animations with the FacialFeatureJitterSuppression parameter of the Tracker API.
- Average enrollment (with FSDK_GetFaceTemplate) speed improvement: 1.5 times.
- Face matching speed improvement:
- Windows: 2.5 to 4 times faster
- macOS: 3 times faster
- Android: 1.5 times faster
- iOS: 1.5 times faster
- Added samples for Visual Studio 2017.
Such dramatic performance improvements were in part possible thanks to the newest development platform. However, we had to drop support for some obsolete operating systems (Windows earlier than XP SP3 or 2003 SP2 and macOS earlier than 10.7). See the Migration chapter in the documentation for details. If you still need your applications to support older versions of the OS, please let us know.
The LiveFacialFeatures, GenderRecognition and ExpressionRecognition samples were updated to be compatible with iOS 11. If your code was based on these samples, visit the Migration chapter to update your code.
FaceSDK 6.2 Release Notes
This version of FaceSDK is about significant performance improvements and more lifelike animations due to the detection of additional facial features. Version 6.2 maintains full backward compatibility with earlier builds, and is a highly recommended update.
- FaceSDK 6.2 detects 70 facial features (up from 66 features in earlier builds). The ‘old’ 66 points retain their numbering for full backwards compatibility. Detecting and tracking more facial features allows making face animations closer to real life.
- Improved tracking of open mouth.
- Improved performance of face detection and facial feature detection.
- Improved performance of retrieving live video in LiveFacialFeatures sample for iOS and Android. The frame rate was significantly increased: our new sample provides solid 30 FPS on an iPhone 6 in landscape mode. This is twice the speed of the earlier release. NOTE: we optimized both the SDK and the sample. In order to achieve this performance increase, you will have to use FaceSDK 6.2 and modify your code.
- Performance improvements allow to significantly reduce jitter of facial animations while maintaining smooth frame rates with the FacialFeatureJitterSuppression parameter of Tracker API.
- Added samples for gender and emotion detection in live video for iOS/Android.
- Added samples for Android Studio and Visual Studio 2015.
- Added the ability to purge faces from Tracker API storage via FSDK_PurgeID.
- Updated Android libraries to be compliant with the latest Google Play requirements.
FaceSDK 6.1 Release Notes
FaceSDK 6.1 adds Facial Expression Recognition, gaining the ability to detect base emotions and determine whether the person has her eyes open and a smile on her face in both stills and videos. The new Expression Recognition sample will help developers quickly integrate new functionality into their apps.
The ability to detect whether the person has her eyes open is invaluable for detecting blinks, allowing taking perfectly timed photos in camera apps or detecting the moment when the person falls asleep. The smile recognition works as a happiness meter, allowing detecting whether the person in the picture or video stream is happy.
Internal tests demonstrate 92% recognition rate for smile detection and nearly 100% rate for open eyes detection.
FaceSDK 6.0.1 Release Notes
FaceSDK 6.0.1 improves compatibility of the LiveRecognition and LiveFacialFeatures samples with iOS 9 and Xcode 7.
FaceSDK 6.0 Release Notes
FaceSDK 6.0 now uses a new, highly optimized Facial Feature Detection and Tracking engine to deliver smooth facial recognition in high-FPS video streams. On desktop platforms, the new SDK now delivers true 60-fps performance on video streams, while video processing on mobile platforms such as Apple iPhone 6 is performed at 16 fps (higher rates are achievable on devices with more powerful platforms). The newly improved engine allows creating real-time augmented reality projects to enable mirror-like performance in applications such as virtual makeup, face morphing, animating or aging.
The new release improves precision and reduces jitter while detecting and tracking facial features in video streams. In addition, FaceSDK 6.0 improves face recognition in video streams, allowing to track faces located near the edge of the frame. Faces that only partially appear in the frame can now be recognized and tracked.
The new SDK once again boosts recognition quality on motion streams, improving False Acceptance Rate by up to 20% when using Tracker API. FaceSDK 6.0 achieved the new low False Rejection Rate of 6.1% on still images (reduced from 6.7% in the last release of the SDK)*.
Finally, the new SDK adds a host of facial recognition samples for both iOS and Android platforms. The new functions enable real-time facial feature detection and tracking on mobile platforms. We’ve also enabled the use of IP-cams on mobile devices, added a sample that searches for similar pictures in Java, and provided buffer functionality (for saving and copying) for Java and .NET development platforms.
* Measured in FRCG test with False Acceptance Rate=0.1%.
FaceSDK 5.0.1 Release Notes
FaceSDK 5.0.1 features a 64-bit build for Apple iOS devices. Version 5.0.1 comes with both 32-bit and 64-bit versions for iOS, allowing building compatible apps for the Apple AppStore. As usual, the update of FaceSDK maintains full code compatibility with previous versions, offering a seamless upgrade path to all who update to the latest build.
FaceSDK 5.0 Release Notes
FaceSDK 5.0 adds a host of revolutionary new features, enabling real-time, motion-based subject identification in video streams with no prior enrollment. iOS and Android platforms are now supported, enabling developers build mobile apps implementing facial identification and recognition. Version 5.0 adds automatic gender recognition with 93% success rate on still images and 97% in motion streams. Finally, face recognition quality on still images is greatly improved with false rejection rate down to only 6.7% (from 9.9% in version 4.0) based on FRGC test.
Version 5.0 constitutes a major upgrade. Updating to this release is highly recommended.Motion-Based Video Recognition
A new revolutionary technology is now available, enabling developers build comprehensive surveillance and security systems, CRM desks and attendance control applications. Implementing secure subject identification based on motion video streams becomes easy with no prior enrollment required.
Many traditional video recognition systems are not truly motion-based. Instead, they are using key frames extracted from the video stream to identify human faces. This approach discards motion-based information and requires a rather complicated training. Enrolling subjects is a lengthy and cumbersome process involving several minutes of posing, and requiring multiple captures from various angles.
FaceSDK 5.0 eliminates the enrollment procedure completely. Every subject appearing in a video is automatically identified and tagged. Enrolling the subject now becomes a simple matter of linking the unique tag to a database record or just putting a name on it.Android and iOS Support
Support for Android and Apple iOS platforms has been added, enabling mobile developers build a variety of apps for the two popular platforms. iOS developers can download Luxand Face Recognition sample app from Apple Store. The Android version is available at Google Play.Automated Gender Recognition
FaceSDK 5.0 can now identify male and female faces on still pictures and motion videos. Still identification rate is 93%, while 97% rate is achievable in video streams.Improved Recognition Quality
FaceSDK 5.0 improves face recognition quality on still images once again, lowering false rejection rate to mere 6.7% (compared to 9.9% in FaceSDK 4.0) based on FRGC test assuming false acceptance rate of 0.1%. This is a 1.48 times improvement compared to previous release.
FaceSDK 4.0 Release Notes
FaceSDK 4.0 improves the speed and accuracy of facial feature detection, with 66 facial features detected in real-time. The 4.0 version supports multi-core processors to boost the performance of face recognition, face detection, and facial feature detection. Support for MJPEG IP cameras has been added.
The increase in speed when using all cores of an Intel i7 processor is 315% for facial feature detection,* 250% for face detection, 130% for face template creation, and 225% for eye detection. The facial feature detection speed is 9.3 frames per second (including face detection using real-time parameters).
Support for Java language has been added. A CImage class for easy manipulation with images on .NET has been added. JPEG file loading was speeded up by 390%.**
More samples were added:
- Lookalikes samples (C#-DB, C#-SQLite, C-DB, C-SQLite) to work with Microsoft SQL and SQLite databases (C#.NET, C++)
- LiveFacialFeatures to detect facial features in real-time (C#.NET, VB.NET, Java, Delphi, C++)
- IPCamera to work with IP cameras (C#.NET, VB.NET, Java, C++)
- FaceTracking, FacialFeatures, and LiveRecognition now feature a Java version.
The LiveRecognition demo application (available on the Welcome screen after the installation) introduces new, more robust usage of FaceSDK recognition algorithms. The Facial Feature Demo application, which tracks a person's facial features in real-time, was added.
Functions to control the number of cores used by FaceSDK were added.
Unicode filename support on Windows was added.
Working with multiple cameras is now thread-safe.
More technical details on migrating from version 3.0 to 4.0 are available in the documentation.
* The speed is measured on an Intel Core i7 930 processor using 8 threads, compared to the same processor using 1 thread.
** The speed is measured on a collection of 1,000 random JPEG files of various resolutions.
FaceSDK 3.0 Release Notes
FaceSDK 3.0 introduces much faster and reliable face recognition. The speed of creating templates for faces was improved by 640% (from 1.3 frames per second to 8.5 frames per second).* The rate of person’s acceptance when lighting conditions vary was improved by 5 times (from 65% to 93%).** This allows to recognize a person in real time regardless of illumination, be it daylight or artificial. Version 3.0 also introduces real-time detection of eye centers.
More samples were added for C# .NET, VB.NET, C/C++, VB6, Delphi and Borland C++ Builder:
- LiveRecognition remembers a person and recognizes her in real time (using a webcam)
- FaceTracking tracks faces in real time
- Lookalikes creates the database of faces and searches for best matches
- FacialFeatures detects facial features on a photo
- Portrait crops a face in command line
Numerous demo applications were added, available in the Welcome screen right after the installation:
- Webcam demo remembers and recognizes persons from a webcam
- Photo demo detects faces on photos
- Panorama demo displays a 3D panorama, tracking the user’s head to change the point of view
The integration with .NET becomes simpler with new FaceSDK.NET assembly. The assembly does not require facesdk.dll and facesdkcam.dll which are already contained inside the assembly.
The facesdkcam.dll file was merged with facesdk.dll, simplifying the deployment and version control for C/C++, Delphi, VB6 and Borland C++ Builder developers.
License Key Wizard was added to simplify requesting an evaluation key from Luxand and pasting it into samples.
Face matching returns a more intuitive value of similarity – a value close to the probability that templates belong to the same person.
The template size was reduced from 92 kilobytes to 16 kilobytes.
Now a camera can be distinguished by its unique ID (device path), which is helpful when multiple cameras of the same manufacturer are used.
A function to mirror the image received from camera was added.
More technical details on migration from version 2.0 to 3.0 are available in the documentation.
* Measured on Intel 2.4 Ghz Processor
** False rejection rate was lowered from 35% to 7% on FERET tests with typical false acceptance rate value = 1%.