That was in 2014. For more on the history of machine learning, check out Roger Parloff 's article at Fortune. big numbers KRAKEN Video Deep Learning Images for High Video Engagement Image reduction is the key to video deep learning. Image analysis is performed through a large number of calculations. Photo: Chase McMichael created the image Think about it: video is a collection of images linked together and played at 30 frames per second. Analyzing a large number of images is a major challenge As humans, we see videos all the time and our brain processes these images in real time.
Getting a machine to do this task on a large scale is not trivial. Machines that process images are an incredible feat, and hair masking service doing this task in real-time video is even more difficult. You have to decipher shapes, symbols, objects and meaning. For robotics and self-driving cars, this is the holy grail. Creating a video image classification system required a slightly different approach. You have to deal with the huge number of unique frames in a video file first to understand what is in the frames. Visual search On September 28, 2016, the seven-member Google Research Team announced that YouTube-8M was built on state-of-the-art deep learning models.
YouTube-8M consists of 8 million YouTube videos, equivalent to 500,000 hours of video, all tagged and there are 4800 Knowledge Graph entities. This is a big deal for the video deep learning space. The scale of YouTube-8M required pre-processing of images to pull the image-level features first. The team used the Inception-V3 image annotation model trained on ImageNet. What makes this so great is that we now have access to a very large video tagging system and Google has done a great job of creating 8M. Google Video Visual Search 8M Stats YouTube 8M top numbers.