Most people don’t know about YouTube-8M, but it has been considered a heavy deal for those working in the field of machine learning. In fact, it is a larger database of labeled video content and it can be used to test out algorithms by programmers. Now, Google has announced that YouTube-8M will receive an important update along with more labels across more its videos, including the audio elements. The search giant is also considering in making the dataset much better using a Kaggle competition. This competition will offer heavy bucks from $100,000 prize pool to teams who build the perfect algorithms for tagging more than 700 thousand new videos by using the dataset training of 8M. Google mentioned on the competition page that the data set was built from more than 450,000 hours of YouTube videos.
These videos include labels from a vocabulary of 4716 classes with an average of 3.4 labels per video. It will present with pre-extracted audio & video features after every second of video. Google also announced that the winning teams will be announced at the YouTube-8M Workshop in the month of July during the Conference of IEEE on Computer Vision & Pattern Recognition. It was said that more than $30 thousand will be awarded per winning team. So, it would be a better chance to attract some eager developers by Google. The search giant also offering some free Google Cloud credits to early participants. But, the competition results will not affect directly to consumers. A software engineer of Google, Paul Natsev pointed out that the learning will be more useful among various types of videos.