ROC SDK 3.16 | 4x Faster Latent Fingerprint Search, Now in SDK

Faster, More Accurate Latent Fingerprint Search

Latent fingerprint matching takes a major leap forward. Searching a gallery of latent prints is now four times faster, templates are 10% smaller, and match accuracy improves at the low false-positive rates forensic examiners rely on. Standard fingerprint capture also templatizes faster, though livescan capture sees modest slowdown in exchange for accuracy gains. This is the speed and precision latent print examination and large-scale forensic search demand. Learn more here.

Know More About Every Person You Detect

ROC SDK 3.16 adds automatic person attribute detection. Alongside every person it detects, the SDK can now predict age, gender, hair, and clothing style and color, adding rich descriptive detail to each detection without an extra processing step. That detail can be used downstream to describe, filter, and investigate people in video feeds, not just locate them.

Sharper, Faster Tattoo Recognition

Tattoo recognition also improves significantly in this release. Error rates are cut in half across the board, search runs 1.5x faster, and templates are smaller, with a modest tradeoff in templatization speeds. Minutiae-only matching has also been upgraded to a new, more accurate model.

Broader Hardware Acceleration

Large galleries can now run first-stage search entirely on the GPU, keeping data resident in device memory for faster search at scale. Hardware acceleration also expands to a new class of devices: Qualcomm-powered Adreno GPUs join NVIDIA as a supported acceleration target. Enabling GPU acceleration no longer requires every algorithm’s model files to be installed, only the ones actually in use.