Tone

In your provided train.log file, the IFRNet_S model has achieved PSNR of 35.46, why you say that you get the PSNR of 34.45?

Some hyperparameters may should be adjusted for IFRNet_S. You can set --lr_end 3e-5, reduce the data augmentatioin probability of [random_resize](https://github.com/ltkong218/IFRNet/blob/1ee657dd4ae6c4b418faed05e347943d0223af47/datasets.py#L10) and [random_rotate](https://github.com/ltkong218/IFRNet/blob/1ee657dd4ae6c4b418faed05e347943d0223af47/datasets.py#L59) and set crop_size=(256, 256) in [random_crop](https://github.com/ltkong218/IFRNet/blob/1ee657dd4ae6c4b418faed05e347943d0223af47/datasets.py#L19). You should...

Thanks for this suggestion! I encourage anyone who can contribute a colab notebook for our project, I will also try to create a colab notebook myself later. Thanks!

Our model can deal with 4K videos, although it may contain larger motion. I think there are two ways to solve this problem: (1) You can use a GPU with...

Welcome anyone to contribute to this project, including conversion methods for onnx. I will also consider this conversion later.

According to your needs, I think you can add the following ensemble inference function to IFRNet.py, IFRNet_L.py and IFRNet_S.py ```python def inference_ensemble(self, img0, img1, embt, scale_factor=1.0): imgt_pred_1 = self.inference(img0, img1,...

I think your suggestion is better than what I have done above. Since directly ensembling the final results will cause blurry texture, while ensembling intermediate optical flow does not have...

I suggest that you can use [opencv-python](https://pypi.org/project/opencv-python/) package and refer to [Python - Extracting and Saving Video Frames](https://stackoverflow.com/questions/33311153/python-extracting-and-saving-video-frames). To answer your question, you can refer to the following steps: 1)...

You can save the [imgt_pred_np](https://github.com/ltkong218/IFRNet/blob/0747fec97b8cb897ab461155978f76f29572d547/demo_2x.py#L21) by imsave('./figures/out_2x.png', imgt_pred_np), where you should import imsave function by from imageio import imsave. For longer image sequences, you can write a loop function and...

The embt is a one-channel conditional input signal for multi frame interpolation. For example, embt = 1/8, 2/8, ..., 7/8 for 8 x frame interpolation.