Salvador Rodríguez

Foundation models like SAM consistently mistake statues, paintings, and reflections for real-world entities—a fatal flaw for 3D reconstruction that is resolved here by routing ambiguous visual matches through a conditional VLM verification layer.