Submitted by Johannes Schönberger.

Submission data

Full nameSGM-Forest
DescriptionSemi-Global Matching (SGM) uses an aggregation scheme to combine costs from multiple 1D scanline optimizations that tends to hurt its accuracy in difficult scenarios. We propose replacing this aggregation scheme with a new learning-based method that fuses disparity proposals estimated using scanline optimization. Our proposed SGM-Forest algorithm solves this problem using per-pixel classification. SGM-Forest currently ranks 1st on the ETH3D stereo benchmark and is ranked competitively on the Middlebury 2014 and KITTI 2015 benchmarks. It consistently outperforms SGM in challenging settings and under difficult training protocols that demonstrate robust generalization, while adding only a small computational overhead to SGM.
Publication titleLearning to Fuse Proposals from Multiple Scanline Optimizations in Semi-Global Matching
Publication authorsJohannes L. Schönberger, Sudipta Sinha, Marc Pollefeys
Publication venueECCV 2018
Programming language(s)C/C++, Python
HardwareIntel Xeon CPU E5-2697, 32GB RAM
Submission creation date28 Feb, 2018
Last edited4 Jul, 2018

High-res multi-view results



Infoallhigh-res
multi-view
indooroutdoorbotani.boulde.bridgedoorexhibi.lectur.living.loungeobserv.old co.statueterrac.
No results yet.

Low-res many-view results



Infoalllow-res
many-view
indooroutdoorlakesidesand boxstorage roomstorage room 2tunnel
No results yet.

Low-res two-view results



Infoalllakes. 1llakes. 1ssand box 1lsand box 1sstora. room 1lstora. room 1sstora. room 2lstora. room 2sstora. room 2 1lstora. room 2 1sstora. room 2 2lstora. room 2 2sstora. room 3lstora. room 3stunnel 1ltunnel 1stunnel 2ltunnel 2stunnel 3ltunnel 3s
two views4.960.322.841.210.6410.236.6411.5510.9810.9413.5911.654.308.944.630.110.040.000.000.050.46