Learning by Analogy: Reliable Supervision from Transformations for Unsupervised Optical Flow Estimation
This work addresses the challenge of accurate optical flow estimation in computer vision, particularly for applications like autonomous driving, by improving unsupervised learning reliability, though it is incremental as it builds on existing unsupervised pipelines.
The paper tackles the problem of unreliable supervision in unsupervised optical flow estimation by introducing a framework that uses transformed data and predictions as self-supervision, achieving state-of-the-art accuracy among deep unsupervised methods and competitive results with supervised methods using fewer parameters.
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.