I2T security camera generates searchable text transcript in real-time
Researchers in California have developed a prototype camera called I2T that can capture video, parse out the background, analyze the action, and then export the content to a text file "that can be searched using simple text search," according to Song-Chun Zhu, lead researcher and professor of statistics and computer science at UCLA. Using a database of over two million images, the system is pretty good at annotating surveillance footage, which generally relies on a static camera. Before the technology can be commercialized, however, they need to significantly expand the database. "If set loose on random images or videos found online," writes Technology Review, the camera would "struggle to perform." If Song-Chun Zhu is interested, we have a library of Greatest American Hero images we can donate to the cause (they're left over from our other blog, GreatestAmericanHeroFanfiction.blogspot.com). Check out the links below to see it in action. [Warning: PDF source link]
