The first 2026 release of 3DiVi Face SDK 3.29 is designed to help businesses speed up performance, strengthen security, and cut operational risks. Here’s what’s new:
Stronger Defense Against Deepfakes, Photos, 2D/3D Masks
New DEEPFAKE_ESTIMATOR v2 cuts detection errors by ~2×, strengthening protection against deepfake scams.
Liveness detection is now more stringent, with the default threshold set to BPCER = 0.01 (down from 0.05), improving resistance to spoofing attacks. The new versions of the 2d_ensemble and 2d_ensemble_light modifications further improve protection accuracy against attacks using 2D and 3D masks.
Liveness detection is now more stringent, with the default threshold set to BPCER = 0.01 (down from 0.05), improving resistance to spoofing attacks. The new versions of the 2d_ensemble and 2d_ensemble_light modifications further improve protection accuracy against attacks using 2D and 3D masks.
Faster Performance, Lower Infrastructure Costs
CUDA 12 support on Linux and expanded parallelization for 1:N search make large-scale face matching significantly faster, reducing server load and infrastructure costs.
The QUALITY_CONTROL block now supports inference on Rockchip NPU, enabling edge devices to run high-accuracy quality assessment locally.
The QUALITY_CONTROL block now supports inference on Rockchip NPU, enabling edge devices to run high-accuracy quality assessment locally.
New Utility Processing Block for Face Normalization
The new Utility Processing Block simplifies obtaining normalized face images for storage as reference photos.
Upgrade to 3DiVi Face SDK 3.29 to deliver safer, and faster face recognition technology for your business.
Upgrade to 3DiVi Face SDK 3.29 to deliver safer, and faster face recognition technology for your business.
