Zijie Wu

Papers

Total Citations

H-Index

About

Zijie Wu is a rising researcher in robotics and computer vision, whose work focuses on enabling long-term autonomous operations through advanced 3D scene understanding. His key contributions lie in developing methods for 3D scene graph prediction and reasoning from RGB-D sequences—a critical step for robots to interpret complex environments over time. In his highly cited 2025 paper, "Hyperrectangle Embedding for Debiased 3D Scene Graph Prediction From RGB Sequences" (41 citations), Wu introduced a novel hyperrectangle embedding approach to address biases in scene graph generation, significantly improving prediction accuracy from sequential data. His follow-up work, "History-Enhanced 3D Scene Graph Reasoning From RGB-D Sequences" (28 citations), further advanced the field by incorporating temporal context to bridge semantic gaps, enabling richer environmental representations. Earlier, Wu demonstrated his versatility with "A Robust Pixel-Wise Prediction Network With Applications to Industrial Robotic Grasping" (17 citations), where he tackled the challenge of 6D pose estimation for textureless industrial parts, enhancing robotic grasping efficiency and robustness. With over 86 total citations and a clear trajectory of impactful, application-driven research, Wu is establishing himself as a key innovator in bridging 3D perception and autonomous robotics.