The KITTI Vision Benchmark Suite

This paper reveals novel geometric principles of
optical flow based on the insight and detailed
definition of non-occlusion. Then, two novel loss
functions are proposed for the unsupervised
learning of optical flow based on the geometric
principles of non-occlusion. Specifically, after
the occlusion part of the images are masked, the
flowing process of pixels is carefully considered
and geometric constraints are conducted based on
the geometric principles of optical flow. First,
neighboring pixels in the first frame will not
intersect during the pixel displacement to the
second frame. Secondly, when the cluster
containing adjacent four pixels in the first frame
moves to the second frame, no other pixels will
flow into the quadrilateral formed by them.
According to the two geometrical constraints, the
optical flow non-intersection loss and the optical
flow non-blocking loss in the non-occlusion
regions are proposed.

All parameters are listed in the paper.

@article{wang2021nccflow,
title={NccFlow: Unsupervised Learning of Optical
Flow With Non-occlusion from Geometry},
author={Wang, Guangming and Ren, Shuaiqi and
Wang, Hesheng},
journal={arXiv preprint arXiv:2107.03610},
year={2021}
}