Closed TNTwise closed 1 month ago
Hello, I have tried this before, but I found that I still don't understand how to improve the performance in a certain scenario. I often find that my expectations contradict user feedback. For example, additional training on anime data does not improve the effect in the corresponding scenario.
Thank you for explaining, again I appreciate the work!
Can u train the model by using a 60 fps source video together with the same video where u remove half of the frames to become a 30 fps video in order for it to learn how to generate the differences? Or model training is harder than that?
In data such as adobe240fps, vimeo90K, the method of data construction is as such. @flowreen
Thank you for your work and time on this project. I am always excited to see what comes next. I was wondering if specified models were planned, for example: one specifically trained for anime interpolation and a separate one trained for realistic interpolation.