Unsupervised Flow-Aligned Sequence-to-Sequence Learning for Video Restoration

Jing Lin, Xiaowan Hu, Yuanhao Cai, Haoqian Wang, Youliang Yan, Xueyi Zou, Yulun Zhang, Luc Van Gool
Proceedings of the 39th International Conference on Machine Learning, PMLR 162:13394-13404, 2022.

Abstract

How to properly model the inter-frame relation within the video sequence is an important but unsolved challenge for video restoration (VR). In this work, we propose an unsupervised flow-aligned sequence-to-sequence model (S2SVR) to address this problem. On the one hand, the sequence-to-sequence model, which has proven capable of sequence modeling in the field of natural language processing, is explored for the first time in VR. Optimized serialization modeling shows potential in capturing long-range dependencies among frames. On the other hand, we equip the sequence-to-sequence model with an unsupervised optical flow estimator to maximize its potential. The flow estimator is trained with our proposed unsupervised distillation loss, which can alleviate the data discrepancy and inaccurate degraded optical flow issues of previous flow-based methods. With reliable optical flow, we can establish accurate correspondence among multiple frames, narrowing the domain difference between 1D language and 2D misaligned frames and improving the potential of the sequence-to-sequence model. S2SVR shows superior performance in multiple VR tasks, including video deblurring, video super-resolution, and compressed video quality enhancement. https://github.com/linjing7/VR-Baseline

Cite this Paper

BibTeX
@InProceedings{pmlr-v162-lin22d, title = {Unsupervised Flow-Aligned Sequence-to-Sequence Learning for Video Restoration}, author = {Lin, Jing and Hu, Xiaowan and Cai, Yuanhao and Wang, Haoqian and Yan, Youliang and Zou, Xueyi and Zhang, Yulun and Van Gool, Luc}, booktitle = {Proceedings of the 39th International Conference on Machine Learning}, pages = {13394--13404}, year = {2022}, editor = {Chaudhuri, Kamalika and Jegelka, Stefanie and Song, Le and Szepesvari, Csaba and Niu, Gang and Sabato, Sivan}, volume = {162}, series = {Proceedings of Machine Learning Research}, month = {17--23 Jul}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v162/lin22d/lin22d.pdf}, url = {https://proceedings.mlr.press/v162/lin22d.html}, abstract = {How to properly model the inter-frame relation within the video sequence is an important but unsolved challenge for video restoration (VR). In this work, we propose an unsupervised flow-aligned sequence-to-sequence model (S2SVR) to address this problem. On the one hand, the sequence-to-sequence model, which has proven capable of sequence modeling in the field of natural language processing, is explored for the first time in VR. Optimized serialization modeling shows potential in capturing long-range dependencies among frames. On the other hand, we equip the sequence-to-sequence model with an unsupervised optical flow estimator to maximize its potential. The flow estimator is trained with our proposed unsupervised distillation loss, which can alleviate the data discrepancy and inaccurate degraded optical flow issues of previous flow-based methods. With reliable optical flow, we can establish accurate correspondence among multiple frames, narrowing the domain difference between 1D language and 2D misaligned frames and improving the potential of the sequence-to-sequence model. S2SVR shows superior performance in multiple VR tasks, including video deblurring, video super-resolution, and compressed video quality enhancement. https://github.com/linjing7/VR-Baseline} }
Endnote
%0 Conference Paper %T Unsupervised Flow-Aligned Sequence-to-Sequence Learning for Video Restoration %A Jing Lin %A Xiaowan Hu %A Yuanhao Cai %A Haoqian Wang %A Youliang Yan %A Xueyi Zou %A Yulun Zhang %A Luc Van Gool %B Proceedings of the 39th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2022 %E Kamalika Chaudhuri %E Stefanie Jegelka %E Le Song %E Csaba Szepesvari %E Gang Niu %E Sivan Sabato %F pmlr-v162-lin22d %I PMLR %P 13394--13404 %U https://proceedings.mlr.press/v162/lin22d.html %V 162 %X How to properly model the inter-frame relation within the video sequence is an important but unsolved challenge for video restoration (VR). In this work, we propose an unsupervised flow-aligned sequence-to-sequence model (S2SVR) to address this problem. On the one hand, the sequence-to-sequence model, which has proven capable of sequence modeling in the field of natural language processing, is explored for the first time in VR. Optimized serialization modeling shows potential in capturing long-range dependencies among frames. On the other hand, we equip the sequence-to-sequence model with an unsupervised optical flow estimator to maximize its potential. The flow estimator is trained with our proposed unsupervised distillation loss, which can alleviate the data discrepancy and inaccurate degraded optical flow issues of previous flow-based methods. With reliable optical flow, we can establish accurate correspondence among multiple frames, narrowing the domain difference between 1D language and 2D misaligned frames and improving the potential of the sequence-to-sequence model. S2SVR shows superior performance in multiple VR tasks, including video deblurring, video super-resolution, and compressed video quality enhancement. https://github.com/linjing7/VR-Baseline
APA
Lin, J., Hu, X., Cai, Y., Wang, H., Yan, Y., Zou, X., Zhang, Y. & Van Gool, L.. (2022). Unsupervised Flow-Aligned Sequence-to-Sequence Learning for Video Restoration. Proceedings of the 39th International Conference on Machine Learning, in Proceedings of Machine Learning Research 162:13394-13404 Available from https://proceedings.mlr.press/v162/lin22d.html.

Related Material