Stereo Matching on Non-Lambertian Generalization
11.4Error Rate (>2px)SMFormer
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| SMFormerVenue=-, Training Dataset=SceneFlow, Evaluation Protocol=Zero-shot, Weight Source=Officially released weights2026.04 | 11.4 | 7 | 4.5 | 4 | 2.8 | |
| DEFOMStereo‡Venue=CVPR’25, Training Dataset=SceneFlow, Evaluation Protocol=Zero-shot, Weight Source=Officially released weights2026.04 | 11.6 | 7.1 | 4.7 | 4.1 | 2.9 | |
| Former-CFNet‡Venue=ECCV’24, Training Dataset=SceneFlow, Evaluation Protocol=Zero-shot, Weight Source=Officially released weights2026.04 | 12 | 7.3 | 5.1 | 4.4 | 3.1 | |
| IGEV-StereoVenue=CVPR’23, Training Dataset=SceneFlow, Evaluation Protocol=Zero-shot, Weight Source=Officially released weights2026.04 | 16.9 | 13.2 | 11.4 | 10.2 | 3.94 | |
| RAFT-StereoVenue=3DV’21, Training Dataset=SceneFlow, Evaluation Protocol=Zero-shot, Weight Source=Officially released weights2026.04 | 17.8 | 13.1 | 10.8 | 9.24 | 3.6 | |
| Selective-IGEVVenue=CVPR’24, Training Dataset=SceneFlow, Evaluation Protocol=Zero-shot, Weight Source=Officially released weights2026.04 | 18.5 | 14.2 | 12.1 | 10.8 | 4.38 | |
| DLNRVenue=CVPR’23, Training Dataset=SceneFlow, Evaluation Protocol=Zero-shot, Weight Source=Officially released weights2026.04 | 18.6 | 14.6 | 12.6 | 11.2 | 3.97 | |
| NRMFVenue=CVPR’24, Training Dataset=SceneFlow, Evaluation Protocol=Zero-shot, Weight Source=Officially released weights2026.04 | 27.1 | 19.1 | 15.4 | 13.2 | 5 |