This clip presents the theory of using technology to not only increase the resolution of digital media, but create details that were not there in the first place. We all know just enlarging a digital image would result in more space taken up, but the image quality doesn’t get better just because you enlarge it.
The method under review here is “Learning Temporal Coherence via Self-Supervision for GAN-based Video Generation” or TecoGAN in short. The machine interprets media and includes the details that it assumes would have been included. Very interesting research happening there. Enjoy!
YouTube: TecoGAN: Super Resolution Extraordinaire