How would you design a feed-forward model to reconstruct a 3D scene from a sparse set of multi-view images for Applied Intuition research workflows?
