貢獻點：

           Construct CNNs which are capable of solving the optical flow estimation problem as a supervised learning task. 

           Train CNN end-to-end to learn predicting the optical flow field from a pair of images

 Propose and compare two architectures: a genetic architecture and another one including a layer that correlates feature vectors at different image locations.

資料庫： KITTI Sintel

演算法：

          While optical flow estimation needs precise per-pixel localization, it also requires finding correspondences between two input images.

          Not only learning image feature representations, but also learn to match them at different locations in the two images. 

框架圖：