With 3DiVi Face SDK 3.31, we’ve introduced a set of updates designed to help developers integrate faster and build more reliable biometric solutions across a wide range of devices.
Key Updates
AI-Agent-Ready Documentation
3DiVi Face SDK now includes the documentation designed specifically for AI agents powered by large language models (LLMs).
It helps AI agents quickly find relevant SDK information and use it when solving development and integration tasks — making it easier for developers to use AI assistants when building applications with Face SDK.
It helps AI agents quickly find relevant SDK information and use it when solving development and integration tasks — making it easier for developers to use AI assistants when building applications with Face SDK.
16 KB Memory Page Support for Google Play
All 3DiVi Face SDK Android libraries now fully support 16 KB memory pages, helping ensure compatibility with current Google Play requirements and avoid potential issues when publishing Android applications.
New Face Tracker Processing Block
The new Face Tracker Processing Block detects and tracks faces in video streams, making it particularly useful for video analytics and access control systems.
Unlike face tracking built into the Face Detector block, the new Face Tracker does not require face detection on every frame.
By reducing the number of detector runs, it can lower computational load and improve video processing efficiency — especially on smartphones and resource-constrained embedded devices.
The new block is already available in Face SDK samples for C++, Java, Kotlin, and Flutter.
Unlike face tracking built into the Face Detector block, the new Face Tracker does not require face detection on every frame.
By reducing the number of detector runs, it can lower computational load and improve video processing efficiency — especially on smartphones and resource-constrained embedded devices.
The new block is already available in Face SDK samples for C++, Java, Kotlin, and Flutter.
Improved Liveness Protection Against Presentation Attacks
A new version (v4) of the 2d_ensemble_light liveness estimation block is now available.
Compared with the previous version, the Attack Presentation Classification Error Rate (APCER) has been reduced by approximately 2x, improving protection against presentation attacks.
Compared with the previous version, the Attack Presentation Classification Error Rate (APCER) has been reduced by approximately 2x, improving protection against presentation attacks.
Support for the Latest Generation of NVIDIA GPUs
Linux inference now supports CUDA 12 and NVIDIA GeForce RTX 50 Series GPUs and other GPUs based on the same architecture.
Power your product with 3DiVi Face SDK — a set of software components for building face recognition solutions across mobile, edge, and server environments.
➟ Explore 3DiVi Face SDK
