Open tjasmin111 opened 1 year ago
How does this model compare with ESRGAN? If we use that tobupscale every single frame. Runtime vs output quality. Hi!
This model is mainly designed for mobile video sr, so the parameter num of the model is very small (tens of k). And the esrgan model is usually a few or tens of M (due to adversarial training) In addition, this model has made a special design for videos (rnn), and esrgan mainly focuses on images.
you can refer to https://arxiv.org/abs/2105.08826 for more details
How does this model compare with ESRGAN? If we use that tobupscale every single frame. Runtime vs output quality.