Latest revision as of 20:09, 1 February 2021

Abstract

Unsupervised learning of optical flow, which leverages the supervision from view synthesis, has emerged as a promising alternative to supervised methods. However, the objective of unsupervised learning is likely to be unreliable in challenging scenes. In this work, we present a framework to use more reliable supervision from transformations. It simply twists the general unsupervised learning pipeline by running another forward pass with transformed data from augmentation, along with using transformed predictions of original data as the self-supervision signal. Besides, we further introduce a lightweight network with multiple frames by a highly-shared flow decoder. Our method consistently gets a leap of performance on several benchmarks with the best accuracy among deep unsupervised methods. Also, our method achieves competitive results to recent fully supervised methods while with much fewer parameters.

Comment: Accepted to CVPR 2020, https://github.com/lliuz/ARFlow

Original document

The different versions of the original document can be found in:

• http://arxiv.org/abs/2003.13045

• http://arxiv.org/pdf/2003.13045

• http://xplorestaging.ieee.org/ielx7/9142308/9156271/09157598.pdf?arnumber=9157598,

http://dx.doi.org/10.1109/cvpr42600.2020.00652

• https://dblp.uni-trier.de/db/conf/cvpr/cvpr2020.html#LiuZHLWTLWLH20,

https://arxiv.org/abs/2003.13045,

https://openaccess.thecvf.com/content_CVPR_2020/papers/Liu_Learning_by_Analogy_Reliable_Supervision_From_Transformations_for_Unsupervised_Optical_CVPR_2020_paper.pdf,

https://openaccess.thecvf.com/content_CVPR_2020/html/Liu_Learning_by_Analogy_Reliable_Supervision_From_Transformations_for_Unsupervised_Optical_CVPR_2020_paper.html,

https://www.arxiv-vanity.com/papers/2003.13045,

https://doi.org/10.1109/CVPR42600.2020.00652,

https://academic.microsoft.com/#/detail/3034896357