AI Computer Vision
3DiVi Inc., founded in 2011, is one of the leading developers of AI and machine learning (ML) technologies for computer vision.
3DiVi News

3DiVi Face Recognition SDK 3.19.0: Advancing Liveness Detection, Image Quality Assurance, and More

The 3DiVi Face Recognition Software Development Kit (SDK) version 3.19.0, released on October 27th, 2023, includes a range of updates and new features designed to improve performance, flexibility, and user experience. Let's discuss the primary updates:

Enhanced Liveness Detection Module

A new version of the Liveness Detection module performs better. Compared to the previous liveness estimator, the updated module has a lower BPCER error rate of 2-3% and a significantly reduced APCER by more than two times for the most frequently occurring attack classes.

Enhanced Quality Assurance Assessment (QAA)

The new version of the Quality Assurance Assessment (QAA) module allows for more efficient filtering of problematic images. Using the new QAA module, you can reduce recognition errors by up to 28% compared to the previous version (*tested on our internal dataset).

Accelerated Template Comparison

Accelerating the template comparison process is a significant improvement, reducing the time needed to work with extensive face databases. The speed of template search through the database has increased by 30%. This acceleration is expected to be a game-changer for enterprises that handle large-scale face recognition tasks.

Detector Configurations for Different Business Cases

Additionally, we've included detector configurations for different business cases. The updated Processing Block API allows users to efficiently use a preconfigured face detector, saving time previously spent selecting optimal parameters.

Extended Processing Block API:

The API includes proven detectors from the previous version, ULD and BLF, as well as several new ones. Additionally, a fitter block has been introduced. Users can access facial recognition blocks in beta mode.

Processing Block demo for the Flutter API:

The newest rollout now includes a Processing Block demo for the Flutter API, further expanding its support for the growing Flutter community. This move will facilitate seamless integration and exploration of 3DiVi's facial recognition technology in Flutter apps.

Updated CUDA and ONNX Versions:

The default versions of CUDA and ONNX used by the SDK have been updated, optimizing the performance and compatibility of the software.

Bug Fixes and Performance Tuning:

Lastly, the 3.19.0 version has addressed some crucial issues to optimize performance, ensuring smoother operation and better resource utilization. A multithreaded mode has been enhanced, underlining 3DiVi's commitment to delivering a more robust and efficient face recognition solution.