Corsight AI facial recognition helps confirm Jesse James photos are genuine

Facial recognition technology equipped with AI is frequently discussed in the context of law enforcement and contemporary fraud. Less so in the case of infamous pistol-shootin’ outlaws from the Wild West.

Corsight AI may yet change that. A release says the company, working in collaboration with academics and historical experts, has been able to confirm using explainable AI that three photos suspected to be of a young Jesse James – notorious bank and train robber of the American West – are indeed him.

“Our technology helps law enforcement and intelligence agencies worldwide identify suspects or missing persons in real time and conduct forensic investigations of photos or videos from crime scenes,” says Rob Watts, President and Chief Strategy Officer of Corsight AI. “As far as I know, this is the first time our recognition technology has been used to identify criminals from the 19th century.”

To make the biometric ID, Corsight examined a collection of historic photos thought to depict James, born in Missouri in 1847 and active in criminal gangs from 1866 until his death by shooting in 1882. Corsight says James was able to evade justice for decades “largely because few people knew what he looked like.”

The hood, the lad and the mugshot: shots of James line up in FRT

The most famous image of James, taken from a wanted poster dated 1881, shows the outlaw as a square-jawed man with slick hair and a stubbly beard – the version often seen in Hollywood Westerns, where James’ legend has grown over time.

Photos dated earlier and presumed to be a youthful James show him with a pistol and a dapper hat during his Confederate guerrilla years, looking a bit like the rapper Eminem. Testing these and others with Corsight’s facial recognition tech “generated extremely high confidence levels for a positive match with the famous WANTED poster portrait.”

“The clear results of the facial recognition analysis we obtained – confirming the authenticity of some photos previously considered questionable – have provided new and surprising insights into the story of Jesse James,” says Warren Getler, a co-author of Rebel Gold, a book about James.

Confidence levels for the candidate photos ranged from 60 to 85. Per the release, facial matching of historic figures “typically does not exceed confidence levels of 85 due to the technical limitations of 19th-century photography and the aging of the original prints, including ‘carte de visite,’ ambrotype, and tintype-style head profiles.”

Which means Corsight’s results were as good as they could be.

Perhaps more importantly, they can be explained.

Explainability pushed in key regulatory frameworks to build trust in AI

Corsight VP of Research and Development Matan Noga says that “with Corsight AI’s unique and patented explainability module, we can explain why the system identified a high match between certain images and a clear mismatch between others.”

With explainability becoming a necessity, that matters. Requiring organizations that use AI to be able to explain, in a way humans can understand, how their algorithms make decisions establishes a greater standard of transparency and trust for interaction with AI.

Yale law professor Shlomit Yaniskiy Ravid notes that “explainability is also emphasized in key regulatory frameworks, including NIST guidelines and the upcoming ISO/IEC 42001 standard, set to take effect in 2025.” It could help “alleviate strict legal frameworks” surrounding the use of facial recognition technologies.

And, as Noga notes, “explainability is also crucial when using AI-based photo or video analysis as evidence in court.”

As for Jesse James, his case is closed… or is it? In the process of making its facial comparisons for a documentary TV series, Corsight discovered that pictures previously considered authentic portraits of James likely aren’t of him.

Article Topics

biometric identification  |  biometrics  |  Corsight  |  criminal ID  |  explainability  |  facial recognition