Back to Basics: Unsupervised Learning of Optical Flow via Brightness Constancy and Motion Smoothness
This addresses the challenge of expensive data labeling in optical flow estimation for computer vision applications, offering an incremental improvement over supervised methods.
The paper tackles the problem of training convolutional networks for optical flow prediction without groundtruth labels by proposing an unsupervised approach using brightness constancy and motion smoothness losses. The result shows that this method outperforms the same network trained with supervision on the KITTI dataset.
Recently, convolutional networks (convnets) have proven useful for predicting optical flow. Much of this success is predicated on the availability of large datasets that require expensive and involved data acquisition and laborious la- beling. To bypass these challenges, we propose an unsuper- vised approach (i.e., without leveraging groundtruth flow) to train a convnet end-to-end for predicting optical flow be- tween two images. We use a loss function that combines a data term that measures photometric constancy over time with a spatial term that models the expected variation of flow across the image. Together these losses form a proxy measure for losses based on the groundtruth flow. Empiri- cally, we show that a strong convnet baseline trained with the proposed unsupervised approach outperforms the same network trained with supervision on the KITTI dataset.