Scientists develop facial recognition software to identify US civil war soldiers.

Washington: Scientists have developed software that uses crowdsourcing to help algorithms identify faces in photos, which could reveal the mysteries of nearly 4 million images of images from the Civil War era. historical record. Kurt Luther, assistant professor at Virginia Tech in the United States, was inspired to develop the Civil War Photo-Sleuth software during a visit to the Heinz History Center exhibition "Pennsylvania Civil War" in Pittsburgh, in Pennsylvania.

There he found a portrait of the time of the Civil War of Oliver Croxton, his great-great-grand-uncle who served in company E of 134th Pennsylvania, dressed in corporal's uniform.

"Historical photos can tell us a lot about not only the history of our family, but also about the historical narrative of the time rather than just reading the event in a history book," said Luther.

The Photo-Sleuth Civil War Project allows users to upload photos, label them with visual cues and link them to profiles of Civil War soldiers with detailed records of military history.

Photo Sleuth's initial reference database contained more than 15,000 portraits of civil war soldiers identified from public domain sources such as the US Military History Institute and other private collections.

More than 600 users sent more than 2,000 civil war photos to the website in the month following their launch, and about half of these photos were not identified.

More than 100 of these unknown photos were associated with specific soldiers and expert analysis revealed that more than 85% of the proposed identifications were probably or definitively correct.

Currently, the database has more than 4,000 registered users and more than 8,000 photos.

"In general, research on group participation like this is a challenge for beginners if users do not have specific knowledge of the subject," said Luther.

"The step-by-step process of marking visual clues and applying state of service search filters makes this detective work more accessible, even for those who do not know the military history of the American Civil War better. "he said.

Person identification may be difficult in larger groups of candidates because of the increased risk of false positives.

Civil War Photo-Sleuth's new approach is based on the analogy of finding a needle in a haystack. The data pipeline has three components related to the haystack: building the haystack, reducing the haystack and finding the needle in the haystack.

Combined, they allow users to identify unknown soldiers while reducing the risk of false positives.

Every time a user uploads a photo to identify it, it is added to the website's digital file or "haystack" of the site, making it available for future searches.

After loading, the user labels the metadata related to the photo, such as the size of the photo or the inscriptions, as well as the visual cues, such as the color of the jacket, the rafters, the shoulder straps, the badge. hat badge.

These tags are linked to search filters to prioritize the most likely matches. For example, a soldier wearing the "hunting horn" badge would suggest potential matches that had been used in infantry while hiding the results of cavalry or artillery.

Next, the site uses advanced facial recognition technology to eliminate very different faces and sort the remaining faces by similarity. The labeling and facial recognition steps reduce the haystack.

Finally, users find the needle in the haystack by exploring in more detail the most likely matches.

A comparison tool with pan and zoom controls helps users to carefully inspect a possible match and, if they choose, connect the previously unknown photo to their new identity and biographical details.

Trace historical photos of the Civil War using facial recognition software such as Photo-Sleuth has many applications that go beyond the identification of historical photos.

Software has the potential to generate new ways of thinking about creating identification systems for people who go beyond facial recognition and take advantage of the complementary strengths of human intelligence and artificial intelligence

Post a Comment