Submitted anonymously.

Submission data

Full nameMASt3R full-improved
DescriptionImproved MASt3R variant with enhanced matching and coarse-to-fine refinement, adapted to the ETH3D SLAM benchmark. RGB-D input is used, and frame-to-frame poses are estimated from MASt3R matches followed by PnP-based pose recovery.
ParametersFrame stride: 15
Pixel tolerance: 5
PnP backend: PoseLib
Reprojection error diagonal ratio: 0.008
PnP max points: 12000
Improved matching: full-improved
Coarse-to-fine refinement enabled
Input resolution follows the MASt3R default patch-compatible resizing.
Programming language(s)Python with PyTorch and CUDA
HardwareNVIDIA GPU, PyTorch/CUDA environment
Submission creation date6 Jun, 2026
Last edited6 Jun, 2026

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
indooroutdoordelivery areaelectroforestplaygroundterrains
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
No results yet.

SLAM results



allboxesboxes darkbuddhacables 4cables 5desk 1desk 2desk changing 2desk dark 1desk dark 2desk global light changesdesk ir lightdinodroneforeground occlusionhelmetkidnap 2lamplarge loop 2large loop 3large non loopmotion 2motion 3motion 4planar 1reflective 2scale changetable 1table 2table 5table 6table global light changestable local light changestable scenetrashbin
MethodInfo
many views4.677.576.955.397.41fail