search

ltkong218 / IFRNet

IFRNet: Intermediate Feature Refine Network for Efficient Frame Interpolation (CVPR 2022)
MIT License
259 stars 23 forks source link

Question #25

Closed dillfrescott closed 1 year ago

dillfrescott commented 1 year ago

How does this method compare to ST-MFNet? Does it yield better results would you say?

https://github.com/danielism97/ST-MFNet

ltkong218 commented 1 year ago

The main differences between IFRNet and ST-MFNet are:

  1. IFRNet takes two frames as input, while ST-MFNet takes four frames as input.
  2. IFRNet uses only Vimeo90K triplet dataset for training, while ST-MFNet uses Vimeo90K septuplet and BVI-DVC datasets for training.
  3. IFRNet is much more efficient and faster than ST-MFNet.

In conclusion, the frame interpolation accuracy of ST-MFNet may be a little higher than IFRNet since it uses additional input frames. However, the accuracy improvement is little when using the same training datasets. Considering the speed and computation, IFRNet achieves better comprehensive VFI performance.

dillfrescott commented 1 year ago

Thank you so much for the info! I appreciate it!