Innovatrics Ranked Among Top Providers of Algorithms for Identification of Latent Fingerprints

Only a handful of global companies have entered the demanding field of identifying latent fingerprints. Innovatrics has also used AI-generated, synthetic fingerprint fragments to augment its algorithm

NIST has published the results of its ELFT (Evaluation of Latent Fingerprint Technology) benchmark. It compares algorithms by companies that provide biometric solutions for criminal investigation in different scenarios.

“Latent fingerprints – i.e. those recovered from a crime scene – pose a number of challenges. They are usually partial, captured in uncontrolled environment, on different backgrounds and with diverse capturing techniques. For us at Innovatrics, it is important to make the identification as convenient as possible, and our results have shown that we rank among the global elite for real-life scenarios,” says Matus Kapusta, Head of the ABIS unit at Innovatrics.

Quality latent fingerprint algorithm is able to reliably and automatically identify newly gathered fingerprints from the crime scene, for example, as part of an ongoing or even a cold case. This opens new possibilities for quickly resolving crimes, identifying suspects in multiple cases, or even returning to those cases that have remained unsolved. The Innovatrics ABIS contains NIST-benchmarked algorithms for the three main biometric modalities: face, fingerprint and iris. The latter two are among the best-performing ones globally. In addition, it contains all the necessary processes for criminal investigation, from enrollment to latent fingerprint enhancement for better identification performance to DNA support.

Apart from the latent fingerprint dataset, Innovatrics has also used an AI-driven approach of generating partial synthetic fingerprints to augment the training dataset and further improve the algorithm. “It’s very difficult to obtain real-world data for this dataset, and synthetic data is a good way to approach this,” explains Jan Lunter, CEO, CTO and Founder of Innovatrics. He adds caution though: “From our experience, synthetic data can be extremely beneficial for training, but one has to be extremely cautious when using them in testing and product validation. Only if testing and validating data are realistic and without bias do we see close to zero risk in using synthetic data in training.”