Neurodactyl, a technology startup founded in Georgia in 2023, archieves world’s top recognition accuracy for border control and visa application datasets in NIST PFT III evaluation. The technology is based on state-of-the-art approaches in neural networks. Besides high recognition accuracy, it also provides extremely fast matching, compact fixed size template and possibility to work with photo images, obtained contactlessly with mobile phone’s camera.
Neurodactyl algorithm demonstrates lowest missing rate for most challenging fingerprint pairs in the datasets with flat fingerprints (“US Visit” and “The Port of Entry/BioVisa Application”): only 0,37% fingerprints were not recognized at false acceptance rate 0.01 and 0.52% at FAR 0.0001, while for recognition by two fingers the missing rate is much lower: 0.06% and 0.08% at FAR 0.01 and 0.0001 correspondingly. Extreme accuracy even for poor quality, “challenging” fingerprints is achieved by implementation the latest approaches in deep learning like GANs and diffusion models. Read full report here.
The algorithm ensures impressive matching speed – from 100 million matching operations per 1 s on one modern CPU like Core i9 and twice as faster on professional CPU like Xeon Gold. Matching can be implemented on GPU with even faster speed – billions of matches per 1 s. It allows to reduce number of servers drastically, which is especially important for large-scale projects with millions of enrollees. The algorithm has small and fixed size biometric template – 512 bytes, regardless of what it is extracted from.
Neurodactyl is a startup specialized in contactless and mobile fingerprint recognition technologies. We develop products and technologies for mobile fingerprints acquisition and recognition, as well as server products for “classic” tasks of verification and identification. Our technologies work with standard scans and photo images, obtained from mobile phones, which opens accurate and reliable fingerprint biometric for Fintech, eKYC, banks and many industries where remote customer’s identification plays a key role.