I can’t pretend to deeply understand what most of this is about. Yesterday on Google’s research blog, it announced the debut of YouTube-8M, a database of 8 million YouTube videos that can be used to improve video search throughout the entire Internet.
I have a feeling, this is something beyond Googling, “new Beyonce video.”
I would like to say here that it’s best to let the experts at Google explain.
YouTube-8M but software engineers Sudheendra Vijayanarasimhan and Paul Natsev, who wrote the post, quickly go deep into the technology woods. “Video analysis provides even more information for detecting and recognizing objects, and understanding human actions and interactions with the world,” they wrote. “Improving video understanding can lead to better video search and discovery, similarly to how image understanding helped re-imagine the photos experience.”
Doing that, with online photographs, is partly what Google Photos does.
When that debuted in May 2015, Google explained that it was reacting to the fact that “humankind” takes a trillion photos a year. It is difficult, and will become more difficult, to keep track of them without some sort of organizing method. The same with video, which is what YouTube-8M is attempting to do for researchers.
The new 8M, the engineers explain, is a “a dataset of 8 million YouTube video URLs (representing over 500,000 hours of video), along with video-level labels from a diverse set of 4,800 Knowledge Graph entities. This represents a significant increase in scale and diversity compared to existing video datasets.”
They note that another exemplary labeled video data set, Sports-1M, has around 1 million YouTube videos and 500 sports-specific classes. Obviously 8M is a helluva lot bigger and more complex.The 8M project only included public videos that have been seen more than 1,000 times and then grouped into 24 verticals. “Arts & Entertainment " defined 2,884,721 videos, more than any other classification. “Finance” videos, the least of the 24. logged only 13,756. For people who work classifying and tagging videos this is big, if esoteric, news.