Release Notes
3.19.2 (02 April 2024)
Bug Fixes and Improvements
- Fixed app_id generation for Android 11+.
- Resolved a licensing error when running multiple Face SDK processes from different users.
- Fixed improper functioning of VideoWorker in Flutter.
- Fixed an error when using online licenses on Android 9.
3.19.1 (14 December 2023)
Bug Fixes and Improvements
- Fixed bug with using Tensorflow libraries on Nvidia Jetson.
- Added new VideoWorker configurations with ssyv detectors.
- Minor fixes for Flutter API.
- Addition of FaceAttributesEstimator to Flutter API.
- Added support for
use_legacy
flag for gender, age, and emotion estimators. - Fixed a bug that caused the malfunction of VideoWorker
store_original_image
flag.
3.17.3/3.18.1 (23 November 2023)
Bug Fixes and Improvements
- Fixed incorrect hash sum of tensorflow libraries for jetpack 4.4, 4.3.
- Fixed excessive RAM consumption when using ONNX models in VideoWorker.
- Fixed memory allocation defect for recognition method 12v50.
- Improved performance of face detectors when used in multithreaded mode (see Capturers use_advanced_multithreading parameter).
3.19.0 (27 October 2023)
What's new
- Added new improved versions of Liveness Detection and Quality Assurance Assessment (QAA) modules.
- Accelerated template comparison procedure. Now it will be faster to work with huge face databases.
- Added Capturer configurations for different business cases. Now the user can get a face detector configured for a specific task without spending time on selecting the best detector and its parameters.
- Processing Block API has been significantly extended. Proven detectors from the old API (ULD and BLF) have been added to the new API, and a number of new ones have been added, for example, face fitter block. Facial recognition blocks are available in beta mode.
- Added Processing Block demo for the Flutter API.
- Updated the default versions of CUDA and ONNX used by Face SDK.
Bug Fixes and Improvements
- Fixed a bug which caused excessive RAM consumption when using the REFA detector.
- Fixed a bug which caused a memory segmentation error when using an empty template_index in the Python API.
- Improved performance of face detectors when used in multithreaded mode.
3.18.0 (13 July 2023)
What's new
- Added CUDA 11 version support
- Added Flutter API documentation
Bug Fixes and Improvements
- Updated Flutter support to 3.3.0 ≤ versions ≤ 3.10.0
3.17.2 (19 June 2023)
Bug Fixes and Improvements
- Fixed a bug due to which matching score was greater than 1 when using recognizer in VideoWorker
- Fixed a bug where 12 versions of recognizers did not run on 32-bit systems
- Updated "score" formula for recognizers
3.17.1 (14 Apr 2023)
What's new
- Added new parameters for recognizers:
num_threads
,inter_op_num_threads
andexecution_mode
- Added ability to pass encoded images to Capturer through Python API
- Added examples of using Processing Blocks in Python
Bug Fixes and Improvements
- Updated "score" formula for recognizers
- Fixed a bug in BLF detector
- Fixed a bug on С# occurred when using Recognizer < 12.x with CUDA acceleration
- Fixed a bug which led to a "segfault" when recreating the FacerecService
- Fixed a bug with a license on Flutter
- Fixed a bug in Python API which led to memory leak
- Fixed visualization bugs in processing_block_demo
- Fixed a bug in Python Samples occurred due to missing library path on Windows
3.17.0 (28 Dec 2022)
What's new
- Added new improved identification algorithms:
method12v30
,method12v50
,method12v100
,method12v1000
(up to 97.4% face recognition accuracy at 1e-8 FAR) - Added ProcessingBlock API for Python
Bug Fixes and Improvements
- Added getFaceVisibilityScore method for RawSample
- Fixed an error for C# API on Android
3.16.1 (31 Oct 2022)
Bug Fixes and Improvements
- Fixed a bug in Quality Assessment Estimation processing block
3.16.0 (8 Sep 2022)
What's new
- Added the new Human Pose Estimator processing block
- Added the new Age Estimation processing block
- Added the new Quality Assessment Estimation processing block
Bug Fixes and Improvements
- Improved algorithm of Age and Gender Estimation
- Improved algorithm of Face Mask detection
- Improved recognizers score computation
3.15.0 (01 Jul 2022)
What's new
- Added the new Object Detector processing block
- Added the new Emotion Estimator processing block
- Added the new Gender Estimator processing block
- Added the new 2D RGB Liveness Estimator processing block
- Added the new algorithm for Emotion Estimation with the "emotion_estimator_v2.xml" configuration
- Added the new algorithm for Gender Estimation with the "age_gender_estimator_v3.xml" configuration
- Added the new algorithm for 2D RGB Liveness Estimation with the "liveness_2d_estimator_v3.xml" configuration
Bug Fixes and Improvements
- Improved algorithm of Quality Estimation
3.14.0 (19 Apr 2022)
What's new
- Added the new Human Body Detector
- Added the new model for estimating state of eyes with the "eyes_openness_estimator_v2.xml" configuration
Bug Fixes and Improvements
- Improved the ULD-detector work
- Fixed bounding boxes for the ULD and BLF detectors
- Fixed an error in VideoWorker that led to memory consumption increase
- Fixed errors in Active Liveness
- Fixed an error in the Flutter Plugin that occurred when online license was used
- Fixed an error in the
get_original_image()
method in Python API
3.13.0 (12 Jan 2022)
What's new
- Added new improved identification algorithm
method11v1000
(up to 99.6% face recognition accuracy) - Added Flutter Support
Bug Fixes and Improvements
- Fixed an error in the CutFaceRawImage method
3.12.0 (22 Oct 2021)
What's new
- Added the new Python demo identify_face.py.
- Added detailed descriptions of license errors.
Bug Fixes and Improvements
- Fixed a memory leak for detectors.
- Fixed errors in c++ demo programs.
- Fixed an error in the setParameter method.
3.11.1 (8 Sep 2021)
Bug Fixes and Improvements
- Fixed templates serialization for 10v methods
3.11.0 (26 Jul 2021)
What's new
- Added new improved identification algorithms:
method10v30
,method10v100
,method10v1000
(up to 99.5% face recognition accuracy) - Added Swift Support for IOS
- Improved Python API
- Added Android 11 support
Bug Fixes and Improvements
- Fixed support of grayscale images for 2D Liveness
- Fixed Access Violation error in java
- Fixed compilation error for IOS
3.10.0 (19 May 2021)
What's new
- Added the Face SDK installer with GUI. Supported platforms:
- Windows 7 or higher
- Ubuntu 16.04 or higher (x64 only)
- Added Python demo programs
- Update VideoWorker STI logic: added timer reset for STI groups, which are tracking (see in Video Stream Processing section)
Bug Fixes and Improvements
- Fixed the error that occurred when the
TemplateCreatedCallback
was called (Python) - Fixed the multiple thread issue in
video_recognition_demo
(Python) - Fixed fail when using VideoWorker on 32-bit systems (Python)
- Fixed fail when using Iris or Active liveness
3.9.0 (29 March 2021)
What's new
- Added Active Liveness
- Added support for a temperature sensor on Telpo
- Added the
RawSample.cutFaceRawImage
method for face cropping and obtaining the result in the decoded form (as an array of pixels) (see Face Cropping) - Added detection confidence level
- Added
FaceAttributesEstimator
for Python - Added the possibility to get an extended set of eye points
- Added estimation of the eyes state
Bug Fixes and Improvements
- Improved the speed of
fda_tracker_capturer
for a large number of people in the frame - Fixed the application for Telpo RGB+IR on Android 9
- Fixed the Segmentation Fault that occurred with the REFA detector in some cases
- Fixed error loading library in Java
3.8.0 (05 Feb 2021)
What's new
- Added display of the used FaceSDK library version
- Added GPU support on Android for
liveness_2d_estimator.xml
- Added GPU support on Jetson (JetPack 4.3/4.4)
- Added the FaceMask Unity sample on Windows and Android
- Added the possibility to estimate the presence of a mask on the face
- Added a new version of RGB Passive Liveness (
liveness_2d_estimator_v2.xml
)
Bug Fixes and Improvements
- Added display of the score parameter for
Liveness2DEstimator
- Added base_angle support for the ULD, BLD, REFA detectors in FDATracker
3.7.1 (18 Jan 2021)
Bug Fixes and Improvements
- Fixed the reading error from
3divi_face_sdk.lic
when using a USB token
v3.7.0 (21 Dec 2020)
What's new
- Added the ULD detector
- Added a beta version of Face SDK VideoEngine JS plugin and demo
Bug Fixes and Improvements
- Fixed fake detections when processing video from multiple streams
v3.6.0 (25 Nov 2020)
What's new
- Added Python API (Windows, Linux)
- Added a new detector providing the highest quality of detection – refa (recommended for expert systems)
- Added GPU usage support on Windows x86 64-bit
- Added a video recognition demo for Telpo Face Recognition Terminals TPS980
- Added a new primitive for evaluating Liveness based on RGB images –
Liveness2DEstimator
(previously available inVideoWorker
)
Bug Fixes and Improvements
- Added support for the AVX2 instruction set (only for Linux x86 64-bit) that can be used to speed up the template creation in 9.x identification methods (on expert systems).
v3.5.0 (21 Oct 2020)
What's new
- Beta version of GPU support for linux-x86-64 and Android
- Added a new module for estimation of age and gender:
age_gender_estimator_v2.xml
. This is a pre-release of a new model.
Bug Fixes and Improvements
- Improved the speed and quality of the 2D Liveness module
- Improved the methods for quality estimation:
quality_estimator_iso.xml
,quality_estimator.xml
v3.4.1 (07 Oct 2020)
Bug Fixes and Improvements
- Updated the tutorial project for iOS
- Fixed demo for Android 11: added the use of scoped storage
v3.4.0 (10 Aug 2020)
What's new
- Added the IR Liveness module (beta)
- Added the BLF detector
- Added the mesh set of points
- Added methods for recognition of masked faces:
method9v30mask_recognizer.xml
method9v300mask_recognizer.xml
method9v1000mask_recognizer.xml
- Added a new improved method for quality estimation:
quality_estimator_iso.xml
v3.3.0.1 (03 June 2020)
Bug Fixes and Improvements
- Added config
video_worker_fdatracker.mask.xml
for masked faces
v3.3.0.0 (27 April 2020)
What's new
- Added new improved identification algorithms:
method9v30
,method9v300
,method9v1000
- Fixed Android demos for Android 10
v3.2.2.4 (02 April 2020)
Bug Fixes and Improvements
- Fixed iOS framework
v3.2.2.3 (27 March 2020)
Bug Fixes and Improvements
- Fixed building C++ examples on Linux
v3.2.2.2 (23 March 2020)
Bug Fixes and Improvements
- Fixed license issue in the android apps
v3.2.2.1 (10 March 2020)
Bug Fixes and Improvements
- Fixed Android 10 support
v3.2.2.0 (17 February 2020)
What's new
- Added iOS support
- Added C# example for 3D-liveness with RealSense
- Added permissions request to Android demo apps
v3.2.1.1 (12 January 2020)
Bug Fixes and Improvements
- Added astra linux 1.5 support
v3.2.1.0 (05 December 2019)
What's new
- Added beta version of 2D-liveness in
VideoWorker
- Added Android Studio samples for android
- Added MS Visual Studio samples for C#
- Added license status logs
- Added match score
v3.2.0.0 (17 October 2019)
What's new
- Added licensing by USB-token
- Added Age, Gender and Emotions to
VideoWorker
- Added API to get license status
- Fixed android signature generation error
v3.1.1.1 (29 July 2019)
Bug Fixes and Improvements
- Added tutorial "Face Recognition in a Video Stream"
v3.1.1.0 (09 July 2019)
What's new
- Added binaries for linux astra 1.11 x86_64 (linux astra 1.5 compatible)
v3.1.0.2 (26 June 2019)
Bug Fixes and Improvements
- Fix memory leak in
VideoWorker
(from v2.4.0.9)
v3.1.0.1 (20 June 2019)
Bug Fixes and Improvements
- Fixed error "0xe078ff45" in
VideoWorker
v3.1.0.0 (10 June 2019)
What's new
Fixed behaviour of
VideoWorker
"max_frames_queue_size" parameter (from v3.0.1.10)Added tutorial "Estimating age, gender, and emotions" (from v3.0.1.9)
Fixed using multiple licenses on the same device (from v3.0.1.8)
Added new improved identification algorithms:
method7.7
,method6.7
,method8.7
.
NOTE:
These methods require results of FDA facial landmarks estimation algorithm,
so if sample was created with different point set, facial landmarks will be
re-estimated by FDA algorithm during Recognizer::processing
.
Also note that result of FDA is more than points position, so samples that are created with
Capturer::manualCapture(..., <vector of points>)
functions will be treated the same way
even if that <vector of points>
contain FDA points set.
So better use FDA capturers configurations, and use RawSample::saveWithoutImage
and FacerecService::loadRawSampleWithoutImage
instead of Capturer::manualCapture
if possible.
- Added new serialization functions:
RawSample::saveWithoutImage
andFacerecService::loadRawSampleWithoutImage
. - Changed format of RawSample serialization. New sdk versions can load old format, old sdk version can't load new format.
- Added new landmarks estimation algorithm - see capturer and
VideoWorker
config files with 'fda' substring.
v3.0.1.7 (7 May 2019)
Bug Fixes and Improvements
- Added tutorial "Displaying Anthropometric Points and Head Rotation Angles"
v3.0.1.6 (24 April 2019)
Bug Fixes and Improvements
- Fixed working of online license on linux arm64
v3.0.1.5 (23 April 2019)
Bug Fixes and Improvements
- Added tutorial "Face Detection and Tracking in a Video Stream"
- Added 'need_stable_results'
VideoWorker
config parameter
v3.0.1.4 (18 April 2019)
Bug Fixes and Improvements
- Fixed precompiled C# demos
- Fixed deadlock in
VideoWorker
(from v2.5.3.7) - Fixed error "0xf4d482c2" which occurred during call of
VideoWorker.resetStream
(from v2.5.3.7)
v3.0.1.3 (07 March 2019)
Bug Fixes and Improvements
- Fixed a bug in the online license when the path to the SDK contains spaces (fixes from v2.5.3.7)
v3.0.1.2 (15 February 2019)
Bug Fixes and Improvements
- Fixed local reference table overflow error in Java api (fixes from v2.5.3.6)
v3.0.1.1 (11 February 2019)
Bug Fixes and Improvements
- Fixed premature disposing of objects in C# api (from v2.5.3.5)
- Fixed using of disposed objects in C# api (from v2.5.3.4)
- Fixed error "use of disposed object - zero pointer
videoWorker_ptr
, code: 0x44cdb4de" incom.vdt.face_recognition.sdk.VideoWorker
(from v2.5.3.3) - Fixed locking at asynchronous call of
VideoWorker.addDepthFrame
andVideoWorker.resetStream/VideoWorker.resetTrackerOnStream
(from v2.5.3.2)
v3.0.1.0 (16 January 2019)
What's new
FacerecNetWrapper.dll
andFacerecNetDemo
have been deprecated- Fixes of examples build (fixes from v2.5.2.4)
- Added
resetStream
method inVideoWorker
(from v2.5.3.1)