3D Scene Understanding
Geometric and semantic scene representations from omnidirectional observations, including 3D occupancy prediction and BEV semantic mapping.
Research in computer vision & robotics
Ph.D. Candidate Hunan University
To address perception bottlenecks faced by robots and autonomous vehicles in complex real-world environments, my research centers on fine-grained visual modeling and understanding of the physical world. I investigate physically grounded geometric and semantic scene representations, as well as controllable generative modeling and learning from synthetic data, with an emphasis on omnidirectional observations.
3D + Visual connected with Generation & Learning
First- or co-first-authored / total
Geometric and semantic scene representations from omnidirectional observations, including 3D occupancy prediction and BEV semantic mapping.
Scene- and motion-conditioned image and video generation, alongside effective learning from synthetic data.
Fine-grained visual understanding across images and videos, spanning scene parsing, object tracking, and multimodal reasoning.
Custom-integrated sensing systems for real-world data collection and perception benchmarking.
A vehicle-mounted system integrating spherical imaging and LiDAR, built to support the Spheriverse dataset and its 3D perception benchmarks.
Video unavailable. Open MP4
A quadruped-mounted system integrating panoramic RGB, thermal, polarization, and LiDAR sensing for the PanoMMOcc dataset and multimodal occupancy benchmark.
Video unavailable. Open MP4
Publications & papers under review
Try a different title, author, or year.
A little background
I am a Ph.D. candidate at Hunan University, advised by Prof. Kailun Yang. I received my M.Eng. from the Karlsruhe Institute of Technology (KIT) and my B.Eng. from China University of Mining and Technology in Xuzhou.
Contact
Research conversations and collaborations are welcome.