Greetings! I am a CS Ph.D. candidate at Peking University, advised by Prof. Peixi Peng. My research interests span scene generation and understanding, with a recent focus on controllable generation and a growing interest in embodied applications of generative models.
Before joining Peking University, I had a pleasant experience as a Research Assistant at The Chinese University of Hong Kong, advised by Prof. Yunhui Liu. I received my master's degree from Shandong University in 2024, where my research focused on scene understanding under the supervision of Prof. Xin Ma. I received my bachelor's degree from the University of Jinan and participated in several robotics competitions at the 1206 Lab, where it all began.
PVWander enables long-horizon panoramic video wandering with an evolving 3D memory, Path-Facing canonicalization, and Timestep-aware memory gating for improved consistency and robustness.
Generative Panoramic video depth estimation with a large-scale RGB-D dataset and Geometry-aware spatiotemporal adaptation
Object semantics are encoded into an inter-class similarity prototype to guide training, improving scene recognition without extra inference cost.
Inter-object discriminative priors enrich scene representations through graph modeling.
Journal extension of SRRM, modeling spatial relations and object co-occurrence for scene recognition.
Semantic region relations for indoor scene recognition.
SHANDONG UNIVERSITY · 2024
基于语义引导的场景识别算法研究
Shandong Provincial Outstanding Master’s Thesis
Exploring semantic guidance for scene recognition through spatial relations, object co-occurrence, discriminative graphs, and similarity prototypes.
2019 · Service Robotics
First Prize · Perfect Score · Team Leader
Watch video · 1:00 · 4× speed ↗2022 · North China Region
First Prize · Team Leader
Watch video · 1:12 ↗Journal Reviewer: IJCV, Pattern Recognition, etc.
Conference Reviewer: ICML, ICLR, etc.
I recently got a Fujifilm X-M5 and plan to give photography a try. Wish me luck!
