What is the difference between super-resolution for animation and super-resolution for photography? .. I compared them.
I compared RealESRGAN’s anime model and photo model
Should I use anime for anime?
Some super-resolution models are optimized for anime.
How is it different from the model for photography?
I actually compared them with the same data.
Easy comparison with image drop
Super-resolution and compare images using CoreML’s model, which allows you to try out the model simply by dropping the image.
You can download the model from GitHub.
https://github.com/john-rocky/CoreML-Models
The size is 4 times larger than the original. Each one is clear, but the difference between the results for animation and for photography is not so obvious.
I will try to zoom in it.
I can hardly understand even if I enlarge it.
How about a photo data?
Zoom in
If you zoom in on this, you can see that the animation version has been crushed.
It has the effect of crushing small artifacts like anime.
Conclusion
If you use animation for super-resolution of photos, the details may be crushed.
It was difficult to make a difference in this data for the super-resolution of anime, but the model for anime seems to have the effect of smoothing.
🐣
I’m a freelance engineer.
Work consultation
Please feel free to contact us with a brief development description.
rockyshikoku@gmail.com
I am making an app that uses Core ML and ARKit.
We send machine learning / AR related information.
