Semi-Supervised Learning of Optical Flow by Flow Supervisor

Woobin Im, Sebin Lee, and Sung-Eui Yoon
TopEuropean Conference on Computer Vision, 2022

Abstract

A training pipeline for optical flow CNNs consists of a pretraining stage on a synthetic dataset followed by a fine tuning stage on a target dataset. However, obtaining ground truth flows from a target video requires a tremendous effort. This paper proposes a practical fine tuning method to adapt a pretrained model to a target dataset without ground truth flows, which has not been explored extensively. Specifically, we propose a flow supervisor for self-supervision, which consists of parameter separation and a student output connection. This design is aimed at stable convergence and better accuracy over conventional self-supervision methods which are unstable on the fine tuning task. Experimental results show the effectiveness of our method compared to different self-supervision methods for semi-supervised learning. In addition, we achieve meaningful improvements over state-of-the-art optical flow models on Sintel and KITTI benchmarks by exploiting additional unlabeled datasets. Code is available at this https URL.

Video

To be updated

BibTex

@inproceedings{Im_2022_ECCV,
author = {Im, Woobin and Lee, Sebin and Yoon, Sung-Eui},
title={Semi-Supervised Learning of Optical Flow by Flow Supervisor},
booktitle = {The European Conference on Computer Vision (ECCV)},
year = {2022}
}