ASFlow: Unsupervised Optical Flow Learning with Adaptive Pyramid Sampling

TL;DR

This paper introduces an unsupervised optical flow estimation method using adaptive pyramid sampling with Content Aware Pooling and Adaptive Flow Upsampling modules, achieving state-of-the-art results on major benchmarks.

Contribution

It proposes novel adaptive pyramid sampling modules that improve feature representation and boundary sharpness in unsupervised optical flow estimation.

Findings

Achieves EPE=1.5 on KITTI 2012

Achieves F1=9.67% on KITTI 2015

Outperforms previous methods by over 13% on key metrics

Abstract

We present an unsupervised optical flow estimation method by proposing an adaptive pyramid sampling in the deep pyramid network. Specifically, in the pyramid downsampling, we propose an Content Aware Pooling (CAP) module, which promotes local feature gathering by avoiding cross region pooling, so that the learned features become more representative. In the pyramid upsampling, we propose an Adaptive Flow Upsampling (AFU) module, where cross edge interpolation can be avoided, producing sharp motion boundaries. Equipped with these two modules, our method achieves the best performance for unsupervised optical flow estimation on multiple leading benchmarks, including MPI-SIntel, KITTI 2012 and KITTI 2015. Particuarlly, we achieve EPE=1.5 on KITTI 2012 and F1=9.67% KITTI 2015, which outperform the previous state-of-the-art methods by 16.7% and 13.1%, respectively.