activity
20162025
most citedPerspective Transformer Nets: Learning Single-View 3D Object Reconstruction without 3D Supervision

311 citations · 979 across the 36 of their papers we have counts for

collaborators
Showing 2019Show all

4 papers · 1 filter