Optical Flow Requires Multiple Strategies (but only one network)
This work addresses optical flow estimation for computer vision applications, offering a novel training approach that improves performance on standard benchmarks.
The paper tackles the optical flow problem by showing it requires multiple strategies depending on image motion, and proposes a metric learning method that selects negative samples based on match nature, achieving state-of-the-art results on KITTI 2012 and KITTI 2015 benchmarks.
We show that the matching problem that underlies optical flow requires multiple strategies, depending on the amount of image motion and other factors. We then study the implications of this observation on training a deep neural network for representing image patches in the context of descriptor based optical flow. We propose a metric learning method, which selects suitable negative samples based on the nature of the true match. This type of training produces a network that displays multiple strategies depending on the input and leads to state of the art results on the KITTI 2012 and KITTI 2015 optical flow benchmarks.