Our new X account is live! Follow @wizwand_team for updates
WorkDL logo mark

Hierarchical Deep Stereo Matching on High-resolution Images

About

We explore the problem of real-time stereo matching on high-res imagery. Many state-of-the-art (SOTA) methods struggle to process high-res imagery because of memory constraints or speed limitations. To address this issue, we propose an end-to-end framework that searches for correspondences incrementally over a coarse-to-fine hierarchy. Because high-res stereo datasets are relatively rare, we introduce a dataset with high-res stereo pairs for both training and evaluation. Our approach achieved SOTA performance on Middlebury-v3 and KITTI-15 while running significantly faster than its competitors. The hierarchical design also naturally allows for anytime on-demand reports of disparity by capping intermediate coarse results, allowing us to accurately predict disparity for near-range structures with low latency (30ms). We demonstrate that the performance-vs-speed trade-off afforded by on-demand hierarchies may address sensing needs for time-critical applications such as autonomous driving.

Gengshan Yang, Joshua Manela, Michael Happold, Deva Ramanan• 2019

Related benchmarks

TaskDatasetResultRank
Stereo MatchingKITTI 2015 (test)
D1 Error (Overall)2.14
144
Stereo MatchingKITTI 2015
D1 Error (All)3.74
118
Disparity EstimationKITTI 2015 (test)
D1 Error (bg, all)1.8
77
Stereo MatchingKITTI 2012 (test)
Outlier Rate (3px, Noc)1.53
76
Stereo MatchingETH3D
bad 1.04.4
51
Stereo MatchingMiddlebury
Bad Pixel Rate (Thresh 2.0)16.5
34
Depth EstimationGated Stereo Day 1.0 (test)
RMSE10.36
19
Depth EstimationGated Stereo Night 1.0 (test)
RMSE12.42
19
Stereo MatchingMiddlebury v3
Average Error2.07
17
Stereo Depth EstimationSQUID zero-shot
Relative Error (Rel)0.9772
16
Showing 10 of 16 rows

Other info

Follow for update