Towards Real-Time Image Enhancement GANs

Fast Video Visual Quality and Resolution Improvement using SR-UNet

“…To tackle variable compression factor, an ensemble of 𝑁 networks is adopting, to avoid mode collapse phenomenon when using a single model. In [10] and [11], Galteri et al address the problem of artifact removal in real-time. Respect to [8], the generator is inspired from the blocks of MobileNetV2, after replacing the standard convolutional layer with lighter depth-wise separable convolutions.…”

Increasing Video Perceptual Quality with GANs and Semantic Coding

“…On the one hand this will "allow" the video call to run smoothly without any delay, on the other hand, the reduced quality in the perceived video will make the user experience, in certain cases, almost unbearable. Recently, solutions to improve image and video quality have been proposed, also running in real-time on tablets and smartphones [14][15][16]. With these algorithms in play it is possible to increase quality and resolution of inbound highly compressed and subsampled videos.…”