Stereo Matching on ETH3D (Threshold Error Breakdown)
0.25Threshold Error > 1px (Noc)MonSter++
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| MonSter++Iters.=32, Latency(s)=7.632026.02 | 0.25 | 0.45 | — | — | 0.18 | — | |
| FoundaStereo(L)Iters.=32, Latency(s)=14.022026.02 | 0.26 | 0.48 | — | — | 0.2 | — | |
| PipStereoIters.=1, Latency(s)=0.442026.02 | 0.35 | 0.67 | — | — | 0.19 | — | |
| MonSterIters.=32, Latency(s)=7.632026.02 | 0.46 | 0.72 | — | — | 0.2 | — | |
| FoundationStereo†Training Set=Unlimited2026.03 | 0.49 | 0.67 | — | 0.17 | — | 0.15 | |
| DEFOMStereoIters.=32, Latency(s)=5.052026.02 | 0.7 | 0.78 | — | — | 0.22 | — | |
| PromptStereoTraining Set=Unlimited2026.03 | 0.79 | 0.97 | — | 0.16 | — | 0.15 | |
| MonSterTraining Set=Unlimited2026.03 | 1.02 | 1.25 | — | 0.19 | — | 0.16 | |
| BridgeDepthTraining Set=Unlimited2026.03 | 1.07 | 1.22 | — | 0.19 | — | 0.18 | |
| IGEVIters.=12, Latency(s)=1.292026.02 | 1.12 | 1.51 | — | — | 0.34 | — | |
| BridgeDepthTraining Set=Scene Flow2026.03 | 1.14 | 1.26 | — | 0.23 | — | 0.22 | |
| SelectiveIGEVIters.=12, Latency(s)=1.612026.02 | 1.23 | 1.56 | — | — | 0.29 | — | |
| PromptStereoTraining Set=Scene Flow2026.03 | 1.3 | 1.56 | — | 0.23 | — | 0.2 | |
| RT-MonSter++Iters.=4, Latency(s)=0.792026.02 | 1.32 | 1.46 | — | — | 0.29 | — | |
| Stereo AnywhereZero-shot=true, Training dataset=SceneFlow2024.12 | 1.43 | 1.66 | 5.29 | 0.24 | — | — | |
| MGStereoTraining Set=Scene Flow2026.03 | 1.59 | 1.88 | — | 0.25 | — | 0.22 | |
| SMoE-StereoTraining Set=Scene Flow2026.03 | 2.08 | 2.44 | — | 0.27 | — | 0.22 | |
| DEFOM-StereoTraining Set=Scene Flow2026.03 | 2.16 | 2.4 | — | 0.35 | — | 0.33 | |
| NS-RAFT-StereoTraining Set=Real-world data without GT2023.03 | 2.23 | 2.94 | — | — | — | — | |
| RAFT-StereoZero-shot=true, Training dataset=SceneFlow2024.12 | 2.24 | 2.59 | 8.78 | 0.25 | — | — | |
| RAFT-StereoTraining Set=Scene Flow2026.03 | 2.29 | 2.61 | — | 0.27 | — | 0.24 | |
| RaftStereoIters.=32, Latency(s)=2.952026.02 | 2.44 | 2.6 | — | — | 0.36 | — | |
| HITNet(L)Iters.=X, Latency(s)=0.442026.02 | 2.79 | 3.11 | — | — | 0.46 | — | |
| MonSter*Training Set=Scene Flow2026.03 | 2.86 | 3.2 | — | 0.35 | — | 0.26 | |
| SMFormer2026.04 | 2.9 | — | — | — | — | — | |
| LoS2026.04 | 3.1 | — | — | — | — | — | |
| RAFT-Stereo2026.04 | 3.3 | — | — | — | — | — | |
| PCVNetZero-shot=true, Training dataset=SceneFlow2024.12 | 3.61 | 4.24 | 14.01 | 0.41 | — | — | |
| IGEV-StereoTraining Set=Scene Flow2026.03 | 3.61 | 4.05 | — | 0.33 | — | 0.29 | |
| NMRFZero-shot=true, Training dataset=SceneFlow2024.12 | 3.66 | 4.34 | 17.15 | 0.42 | — | — | |
| GREAT-IGEVTraining Set=Scene Flow2026.03 | 3.8 | 4.47 | — | 0.37 | — | 0.28 | |
| Selective-RAFTZero-shot=true, Training dataset=SceneFlow2024.12 | 3.81 | 4.36 | 10.23 | 0.34 | — | — | |
| RT-IGEV++Iters.=6, Latency(s)=0.392026.02 | 3.81 | 4.45 | — | — | 0.66 | — | |
| IGEV-StereoZero-shot=true, Training dataset=SceneFlow2024.12 | 3.86 | 4.3 | 12.65 | 0.38 | — | — | |
| IGEVStereo2026.04 | 4.1 | — | — | — | — | — | |
| UCFNet_pretrain2026.04 | 4.8 | — | — | — | — | — | |
| AANetIters.=X, Latency(s)=0.482026.02 | 5.01 | 5.41 | — | — | 0.58 | — | |
| PCWNet2026.04 | 5.2 | — | — | — | — | — | |
| GMStereoZero-shot=true, Training dataset=SceneFlow2024.12 | 5.58 | 6.22 | 19.97 | 0.42 | — | — | |
| FastACVNet+Iters.=X, Latency(s)=0.272026.02 | 5.62 | 6 | — | — | 0.63 | — | |
| Selective-IGEVZero-shot=true, Training dataset=SceneFlow2024.12 | 5.71 | 6.42 | 18.71 | 1.73 | — | — | |
| CFNet2026.04 | 5.8 | — | — | — | — | — | |
| IINetIters.=X, Latency(s)=0.292026.02 | 5.94 | 6.21 | — | — | 0.57 | — | |
| Former-PSMNet (SAM)Pre-trained VFM=true2026.04 | 6.4 | — | — | — | — | — | |
| NS-PSMNetTraining Set=Real-world data without GT2023.03 | 11 | 11.69 | — | — | — | — | |
| PSMNetZero-shot=true, Training dataset=SceneFlow2024.12 | 18.62 | 19.75 | 42.05 | 0.94 | — | — | |
| CoExIters.=X, Latency(s)=0.172026.02 | 19.78 | 20.15 | — | — | 5.95 | — | |
| LightStereo(S)Iters.=X, Latency(s)=0.132026.02 | 21.82 | 22.81 | — | — | 8.15 | — | |
| DLNRZero-shot=true, Training dataset=SceneFlow2024.12 | 22.94 | 23.12 | 26.93 | 9.89 | — | — | |
| ELFNetZero-shot=true, Training dataset=SceneFlow2024.12 | 24.5 | 25.61 | 46.06 | 5.65 | — | — | |
| RAFT-StereoTraining Set=SceneFlow with GT, Protocol=Standard2023.03 | — | 3.28 | — | — | — | — |