Object-Centric Representation Learning for Enhanced 3D Semantic Scene Graph Prediction | Read Paper on Bytez