PhD student, Ecole Nationale des Ponts et Chausees
1 paper at NeurIPS 2025
A new pre-training method for two-view pose estimation, which is classifying each pixel as co-visible, occluded, or outside field-of-view in the other view.
