-
Notifications
You must be signed in to change notification settings - Fork 424
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Training strategies of the three models provided #79
Comments
Hello, we train RIFE and RIFE2F using the same training strategies. RIFE_HD is designed for better visual effect on HD videos, so I add scale augmentation for Vimeo90K. (1.5x 2x 4x 6x randomly) |
if the model input size is A, do you mean 1.5xA 2x A 4xA 6xA randomly crop the data, and then resize them to A? |
Yes. You should rescale and multiply the optical flow label as well. |
hello, the size of the image of vimeo90 is 256x448, the crop size that you set in project is 224x224(model input size A). according to the above discussion,how to crop patch of size 6xA. do you use smaller A? such as 42x42? |
Resize first and then crop. |
do you train the RIFE RIFE2F RIFE_HD using the same training strategies such as learning_rate、optimizer、training epoch and so on?
The text was updated successfully, but these errors were encountered: