AMNet: Deep Atrous Multiscale Stereo Disparity Estimation Networks

TL;DR

This paper introduces AMNet, a deep multiscale stereo disparity estimation network utilizing atrous convolutions and an extended cost volume, achieving state-of-the-art accuracy on multiple benchmarks.

Contribution

It presents a novel deep learning architecture with multiscale feature extraction and a foreground-background aware variant, trained via an iterative multitask method.

Findings

Achieves high accuracy on Middlebury, KITTI, and Sceneflow benchmarks.

Outperforms previous methods in stereo disparity estimation.

Demonstrates effective multiscale contextual information aggregation.

Abstract

In this paper, a new deep learning architecture for stereo disparity estimation is proposed. The proposed atrous multiscale network (AMNet) adopts an efficient feature extractor with depthwise-separable convolutions and an extended cost volume that deploys novel stereo matching costs on the deep features. A stacked atrous multiscale network is proposed to aggregate rich multiscale contextual information from the cost volume which allows for estimating the disparity with high accuracy at multiple scales. AMNet can be further modified to be a foreground-background aware network, FBA-AMNet, which is capable of discriminating between the foreground and the background objects in the scene at multiple scales. An iterative multitask learning method is proposed to train FBA-AMNet end-to-end. The proposed disparity estimation networks, AMNet and FBA-AMNet, show accurate disparity estimates and advance the state of the art on the challenging Middlebury, KITTI 2012, KITTI 2015, and Sceneflow stereo disparity estimation benchmarks.