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.

We analyse the convergence of one-hidden-layer ReLU networks trained by gradient flow on n data points, when the input lies in very high dimension.