List your company
ResearchFormerly WizWand

Monocular Depth Estimation on ScanNet++ (test)

0.0774RMSE2K Retrofit
0.0500560.2346280.41920.603772Aug 6, 2025Sep 12, 2025Oct 20, 2025Nov 27, 2025Jan 3, 2026Feb 10, 2026Mar 20, 2026
Updated 7mo ago

Evaluation results

MethodLinks
2K Retrofitzero-shot=false2026.030.0774—1.4698.25
PromptDAzero-shot=false2026.030.0829—1.7597.81
2K Retrofitzero-shot=true2026.030.0853—1.9798.01
PromptDAzero-shot=true2026.030.0966—2.2497
MSPFzero-shot=false2026.030.0975—2.2696.74
PatchRefinerzero-shot=true2026.030.0976—2.3796.83
Depth Promptingzero-shot=true2026.030.0983—2.4296.57
DepthAnything v2*zero-shot=false, finetuned=true2026.030.101—3.7194.37
DepthAnything v2zero-shot=true2026.030.1145—4.0294.04
Marigoldzero-shot=true2026.030.1412—6.0387.18
BPNetzero-shot=true2026.030.1963—14.1993.57
UniDepth + Calibration TokensExperiment=+ Calibration Tokens, Model=UniDepth, Train Dataset=Mix 200K, Zero-shot=true2025.080.24476.6——
DepthAnyCameraExperiment=Comparisons, Model=DepthAnyCamera, Train Dataset=Indoor 670K, Zero-shot=true2025.080.27576.1——
UniDepthExperiment=Baseline, Model=UniDepth, Train Dataset=Mix 16M, Zero-shot=true2025.080.27972——
FoVA-DepthExperiment=Comparisons, Model=FoVA-Depth, Train Dataset=Indoor 190K, Zero-shot=true2025.080.28554.8——
MiDaS + Calibration TokensExperiment=+ Calibration Tokens, Model=MiDaS, Train Dataset=Mix 200K, Zero-shot=true2025.080.44656.9——
MiDaSExperiment=Baseline, Model=MiDaS, Train Dataset=Mix 1.4M, Zero-shot=true2025.080.50656.3——
DepthAnything + Calibration TokensExperiment=+ Calibration Tokens, Model=DepthAnything, Train Dataset=Mix 200K, Zero-shot=true2025.080.60750.6——
DepthAnythingExperiment=Baseline, Model=DepthAnything, Train Dataset=Mix 63.5M, Zero-shot=true2025.080.73146.3——
DepthAnyCameraExperiment=Comparisons, Model=DepthAnyCamera, Train Dataset=Mix 200K, Zero-shot=true2025.080.76125.5——