A Perceptual Quality Metric for Video Frame Interpolation

Paper | Data | Video | Poster

PyTorch implementation of A Perceptual Quality Metric for Video Frame Interpolation.

A Perceptual Quality Metric for Video Frame Interpolation,
Qiqi Hou ¹, Abhijay Ghildyal ¹, Feng Liu ¹,
¹Portland State University

in European Conference on Computer Vision (ECCV) 2022.

As video frame interpolation results often exhibit unique artifacts, existing quality metrics sometimes are not consistent with human perception when measuring the interpolation results. Some recent deep learning-based perceptual quality metrics are shown more consistent with human judgments, but their performance on videos is compromised since they do not consider temporal information. In this project, we present a dedicated perceptual quality metric for measuring video frame interpolation results.

Video

Network

Requirements

System: Ubuntu

Pytorch version: >= Pytorch1.10

GPU memory: >= 12G

Test on BVI-VFI dataset

1. Download the BVI-VFI dataset. Set the path in the function get_dataset_dir in utils.py

    elif dataset == 'bvivfi':
        datadir = 'BVI-VFI_DATASET_PATH'

2. Run the testing codes

python test_bvivfi_fast.py --model=multiscale_v33 --expdir=./exp/eccv_ms_multiscale_v33/ --testset=bvivfi

Training

1. Download the VFIPS dataset (Please contact with qiqi.hou2012@gmail.com for it.). Set the path in the function get_dataset_dir in utils.py

    if dataset == 'vfips':
        datadir = 'VFIPS_DATASET_PATH'

2. Run the training codes

python train.py --model=multiscale_v33 --expdir=./exp/eccv_ms_multiscale_v33/

Citation

If you find this code is helpful, please consider to cite our paper.

@inproceedings{hou2022vfips,
  title={A Perceptual Quality Metric for Video Frame Interpolation},
  author={Qiqi Hou, Abhijay Ghildyal, Feng Liu},
  year={2022},
  booktitle = {European Conference on Computer Vision},
}

If you find any bugs of the code, feel free to send me an email: qiqi.hou2012@gmail.com. You can find more information in my homepage.

Acknowledgments

This work was made possible in part thanks to Research Computing at Portland State University and its HPC resources acquired through NSF Grants 2019216 and 1624776. Source frames are used under a Creative Commons license from Youtube users Ignacio, Scott, Animiles, H-Edits, 3 Playing Brothers, TristanBotteram, popconet, and billisa.