Stereo Matching on Middlebury half resolution (test)
1.12Threshold Error RateFoundationStereo
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| FoundationStereoTraining Data=Million-scale, MACs (G)=12824, Evaluation Protocol=Zero-shot2025.11 | 1.12 | 0.37 | |
| Selective-IGEV‡Training Data=Million-scale, MACs (G)=3619, Evaluation Protocol=Zero-shot2025.11 | 7.5 | — | |
| Lite Any StereoTraining Data=Million-scale, MACs (G)=33, Evaluation Protocol=Zero-shot2025.11 | 7.51 | 0.94 | |
| StereoAnything-L†Training Data=Million-scale, MACs (G)=84, Evaluation Protocol=Zero-shot2025.11 | 9.82 | 1.21 | |
| BANet-2D*Training Data=Million-scale, MACs (G)=36, Evaluation Protocol=Zero-shot2025.11 | 10.05 | 1.34 | |
| FC-GANetTraining data=SceneFlow + Asymmetric Augmentation2022.03 | 10.2 | — | |
| LightStereo-M*Training Data=Million-scale, MACs (G)=33, Evaluation Protocol=Zero-shot2025.11 | 10.85 | 1.51 | |
| FC-DSMNetTraining data=SceneFlow2022.03 | 12 | — | |
| Lite Any StereoTraining Data=Scene Flow, MACs (G)=33, Evaluation Protocol=Zero-shot2025.11 | 13.13 | 1.6 | |
| GANetTraining data=SceneFlow + Asymmetric Augmentation2022.03 | 13.5 | — | |
| DSMNetTraining data=SceneFlow2022.03 | 13.8 | — | |
| Lite-CREStereo++Training Data=Scene Flow, MACs (G)=101, Evaluation Protocol=Zero-shot2025.11 | 14.91 | 3.32 | |
| FC-PSMNetTraining data=SceneFlow + Asymmetric Augmentation2022.03 | 15.1 | — | |
| CFNetTraining data=SceneFlow + Asymmetric Augmentation2022.03 | 15.3 | — | |
| STTRTraining data=SceneFlow + Asymmetric Augmentation2022.03 | 15.5 | — | |
| PSMNetTraining data=SceneFlow + Asymmetric Augmentation2022.03 | 15.8 | — | |
| FC-GANetTraining data=SceneFlow2022.03 | 16 | — | |
| LightStereo-MTraining Data=Scene Flow, MACs (G)=33, Evaluation Protocol=Zero-shot2025.11 | 16.99 | 2.06 | |
| LightStereo-LTraining Data=Scene Flow, MACs (G)=84, Evaluation Protocol=Zero-shot2025.11 | 17.23 | 2.88 | |
| FC-PSMNetTraining data=SceneFlow2022.03 | 18.3 | — | |
| MS-GCNetTraining data=SceneFlow2022.03 | 18.5 | — | |
| FastACVTraining Data=Scene Flow, MACs (G)=72, Evaluation Protocol=Zero-shot2025.11 | 19.61 | 4.66 | |
| MS-PSMNetTraining data=SceneFlow2022.03 | 19.9 | — | |
| GANetTraining data=SceneFlow2022.03 | 20.3 | — | |
| FC-GWCNetTraining data=SceneFlow2022.03 | 21 | — | |
| SGMTraining data=N/A2022.03 | 25.2 | — | |
| MobileStereoNet-3DTraining Data=Scene Flow, MACs (G)=564, Evaluation Protocol=Zero-shot2025.11 | 25.26 | 4.61 | |
| CoEXTraining Data=Scene Flow, MACs (G)=54, Evaluation Protocol=Zero-shot2025.11 | 26.42 | 4.9 | |
| BANet-2DTraining Data=Scene Flow, MACs (G)=36, Evaluation Protocol=Zero-shot2025.11 | 26.79 | 6.96 | |
| PSMNetTraining data=SceneFlow2022.03 | 26.9 | — | |
| FastACV+Training Data=Scene Flow, MACs (G)=85, Evaluation Protocol=Zero-shot2025.11 | 27.34 | 7.16 | |
| BANet-3DTraining Data=Scene Flow, MACs (G)=78, Evaluation Protocol=Zero-shot2025.11 | 28.79 | 8.05 | |
| GWCNetTraining data=SceneFlow2022.03 | 34.2 | — | |
| MobileStereoNet-2DTraining Data=Scene Flow, MACs (G)=127, Evaluation Protocol=Zero-shot2025.11 | 37.98 | 7.54 | |
| PatchMatchTraining data=N/A2022.03 | 38.6 | — | |
| CostFilterTraining data=N/A2022.03 | 40.5 | — |