twinsyssy1018 / Super-SloMo

PyTorch implementation of Super SloMo by Jiang et al.

Super-SloMo

PyTorch implementation of "Super SloMo: High Quality Estimation of Multiple Intermediate Frames for Video Interpolation" by Jiang H., Sun D., Jampani V., Yang M., Learned-Miller E. and Kautz J. [Project] [Paper]

Results

Results on UCF101 dataset using the evaluation script provided by paper's author. The get_results_bug_fixed.sh script was used. It uses motions masks when calculating PSNR, SSIM and IE.

Method	PSNR	SSIM	IE
DVF	29.37	0.861	16.37
SepConv - L_1	30.18	0.875	15.54
SepConv - L_F	30.03	0.869	15.78
SuperSloMo_Adobe240fps	29.80	0.870	15.68
pretrained mine	29.77	0.874	15.58
SuperSloMo	30.22	0.880	15.18

Prerequisites

This codebase was developed and tested with pytorch 0.4.1 and CUDA 9.2.

Training

Preparing training data

In order to train the model using the provided code, the data needs to be formatted in a certain manner.
The create_dataset.py script uses ffmpeg to extract frames from videos.
For adobe240fps, download the dataset, unzip it and then run the following command

python data\create_dataset.py --ffmpeg_dir path\to\ffmpeg --videos_folder path\to\adobe240fps\videoFolder --dataset_folder path\to\dataset --dataset adobe240fps

Training

In the train.ipynb, set the parameters (dataset path, checkpoint directory, etc.) and run all the cells.

Tensorboard

To get visualization of your training, you can run tensorboard from the project directory using the command:

tensorboard --logdir logs --port 6007

Evaluation

Pretrained model

You can download the pretrained model trained on adobe240fps dataset here.

More info TBA

To-Do's:

Task	Status
Add evaluation script for UCF dataset	TBD
Add getting started guide	TBD
Add video converter script	In progress