Attention-based motion aggregation concepts have recently shown their usefulness in
optical flow estimation, in particular when it comes to handling occluded regions. However,
due to their complexity, such concepts have been mainly restricted to coarse-resolution
single-scale approaches that fail to provide the detailed outcome of high-resolution multi-
scale networks. In this paper, we hence propose CCMR: a high-resolution coarse-to-fine
approach that leverages attention-based motion grouping concepts to multi-scale optical …

In the following, we provide additional details on our approach. Regarding the architecture,
we further detail on the improved feature consolidation unit and how it is embedded in the
whole feature extractor. We then show additional pre-training generalization results,
comparing our approach to its baselines. After that we give additional details on the memory
advantages of our approach compared to a simple coarse-to-fine GMA variant. We then
describe the validation set on KITTI that we used for delineate the architectural from the fine …