Sim2SG: Generating Sim-to-Real Scene Graphs for Transfer Learning

Originally published at: Sim2SG: Generating Sim-to-Real Scene Graphs for Transfer Learning | NVIDIA Developer Blog

Scene graphs (SGs) in both computer vision and computer graphics are an interpretable and structural representation of scenes. A scene graph summarizes entities in the scene and plausible relationships among them. SGs have applications in the fields of computer vision, robotics, autonomous vehicles, and so on. Current SG-generation techniques rely on the limited availability of…

SIM2SG, scene graph generation for transfer learning, closes the Domain Gap between synthetic and real data - check it out and tell us what you think.

Recently, scene graph (SG) generation has gained a lot of traction. Current SG generation techniques, on the other hand, depend on the costly and restricted availability of labelled datasets. Since labels are basically free, synthetic data is a viable alternative. However, due to the domain distance, neural network models trained on synthetic data do not perform well on real data.