NAVA—part of MTVA, the National Audiovisual Archive of Hungary—acts to preserve the country’s rich database of twentieth-century imagery and audio heritage. Microsoft Azure Cognitive Services has proven the only tool to help do what the team needs: accurately classify millions of images that will serve as the country’s long-term collective memory bank.

An enormous and socially vital task

A storehouse for over 140 years of the republic’s public television, radio, and news agency output, MTVA, Hungary’s national media archives, curates a vast library of social legacy; the library contains 20 million printed news stories dating back to 1881 and 350,000 newsreels on a mix of technology, including 16mm and 35mm films as well as Quadruplex and BetaSP videotapes. Properly preserving this valuable resource for citizens and researchers presents numerous problems; in 2006, the archives set up a technology research team, NAVA, to automate and digitize the collection.

Work has progressed since then, with 10 percent of news stories and 30 percent of images electronically stored. But as NAVA General Manager Lipót Répászky explains, the size of the task meant that any technology help for the staff would be beneficial, as the team’s small 100-person staff can’t work at the pace and scale required for high-speed archiving. “One person cannot know who every public figure is, but we wondered if properly trained AI could,” says Répászky.

An 80 percent time reduction

Répászky found an answer: Microsoft Azure Cognitive Services, which could apply intelligent support to archiving. According to Répászky, “We have a target of 2.5 million images a year fully digitized, but really were only achieving 40 percent of that goal, so we needed the best technology support for our people we could find.”

After an extensive market evaluation, Microsoft AI emerged as the only practical way to do what NAVA wanted. “Cognitive Services was the only solution we found that gave us the scale and the ability to integrate with any other software we needed, and was very user-friendly in terms of building services,” Répászky confirms. “The AI is also so accurate, it can recognize 90 percent of features in a photograph instantly.”

An AI solution helps by very rapidly doing a lot of the basic classification work for image filing—filling in important fields, including well-known people and prominent buildings. The reduction in archivist workload is significant, according to Répászky. “What used to take a person 10 minutes without AI now takes more like two with AI.”

Cognitive Services was the only solution we found that gave us the scale and the ability to integrate with any other software we needed, and was very user-friendly in terms of building services.

– Lipót Répászky: General Manager

NAVA

Menu