Timezone: »

 
Poster
Semi-Supervised Learning for Optical Flow with Generative Adversarial Networks
Wei-Sheng Lai · Jia-Bin Huang · Ming-Hsuan Yang

Mon Dec 04 06:30 PM -- 10:30 PM (PST) @ Pacific Ballroom #90
in Posters Mon »

Convolutional neural networks (CNNs) have recently been applied to the optical flow estimation problem. As training the CNNs requires sufficiently large ground truth training data, existing approaches resort to synthetic, unrealistic datasets. On the other hand, unsupervised methods are capable of leveraging real-world videos for training where the ground truth flow fields are not available. These methods, however, rely on the fundamental assumptions of brightness constancy and spatial smoothness priors which do not hold near motion boundaries. In this paper, we propose to exploit unlabeled videos for semi-supervised learning of optical flow with a Generative Adversarial Network. Our key insight is that the adversarial loss can capture the structural patterns of flow warp errors without making explicit assumptions. Extensive experiments on benchmark datasets demonstrate that the proposed semi-supervised algorithm performs favorably against purely supervised and semi-supervised learning schemes.

Keywords: Semi-Supervised Learning · Computer Vision · Adversarial Networks
[ Paper] 

Author Information

Wei-Sheng Lai (University of California, Merced)
Jia-Bin Huang (Virginia Tech)
Ming-Hsuan Yang (UC Merced / Google)

More from the Same Authors