NeRF-Supervised Deep Stereo

Fig. 1 . Zero-Shot Generalization Results. On top, predictions by trained with our approach on user-collected images, without using any synthetic datasets, ground-truth depth or (even) real stereo pairs. At the bottom, a zoom-in over the Backpack disparity map, showing an unprecedented level of detail compared to existing strategies not using ground-truth [3, 23, 79] trained with the same data.