Abstract
We propose a new self-supervised approach to image feature learning frommotion cue. This new approach leverages recent advances in deep learning in twodirections: 1) the success of training deep neural network in estimatingoptical flow in real data using synthetic flow data; and 2) emerging work inlearning image features from motion cues, such as optical flow. Building onthese, we demonstrate that image features can be learned in self-supervision byfirst training an optical flow estimator with synthetic flow data, and thenlearning image features from the estimated flows in real motion data. Wedemonstrate and evaluate this approach on an image segmentation task. Using thelearned image feature representation, the network performs significantly betterthan the ones trained from scratch in few-shot segmentation tasks.