Hi there, I am Fan Li (李帆), a Ph.D. candidate at the School of Artificial Intelligence (formerly the Unmanned System Research Institute), Northwestern Polytechnical University (NPU). My research focuses on the intersection of unmanned systems and computer vision, dedicated to overcoming key visual perception bottlenecks faced by intelligent unmanned platforms operating in real-world, complex, and dynamic environments.
Specifically, my research interests include:
-
Generative-driven next-generation perception paradigms
-
Generalization of unmanned systems’ perception in complex and dynamic environments
-
Robust visual understanding in open-world scenarios
-
Multi-modal 3D scene understanding and representation learning
My research aims to enable resilient and trustworthy visual perception and scene understanding for unmanned systems operating in complex, dynamic, and uncertain environments. This work strives to significantly expand the operational envelope of unmanned systems, facilitating their scaling from controlled testbeds to unstructured and real-world scenarios.
I am always open to collaborations and interdisciplinary discussions. If you are interested in exploring potential research avenues or discussing related topics, please contact me at lifan.messages@gmail.com.
🔥 News
- 2025.09: 🎉 A paper on 3D semantic segmentation was accepted to NeurIPS 2025 Main Track.
- 2025.07: 🎉 Our ICCV 2025 paper was selected as a Highlight.
- 2025.06: 🎉 A paper on open-world semantic segmentation was accepted to ICCV 2025.
- 2025.05: 🎉 A paper on domain generalization was accepted to ICML 2025 and was awarded as a Spotlight.
- 2025.03: 🎉 A paper on domain adaptation was accepted to ICME 2025.
📝 Publications
Fan Li, Xuan Wang, Min Qi, Zhaoxiang Zhang, Yuelei Xu
Accepted at ICML 2025 (Spotlight💡)
- Developed an agent-query–driven framework for domain-generalized semantic segmentation, in which learnable agent queries act as an interface to inject diffusion-model priors into visual perception networks, thereby improving cross-domain generalization.
Fan Li, Xuanbin Wang, Xuan Wang, Zhaoxiang Zhang, Yuelei Xu
Accepted at ICCV 2025 (Highlight💡)
- Developed the first denoising-based learning framework for visual perception, which treats images as noisy labels and acquires semantic knowledge during the iterative denoising process of diffusion models, thereby enhancing fine-grained scene understanding in open-world scenarios for vision–language models.
No Object Is an Island: Enhancing 3D Semantic Segmentation Generalization with Diffusion Models
Fan Li, Xuan Wang, Xuanbin Wang, Zhaoxiang Zhang, Yuelei Xu
Accepted at NeurIPS 2025 Main Track
- Developed the first diffusion-based cross-domain 3D semantic segmentation framework, leveraging pretrained 2D image diffusion priors to guide 3D models in establishing semantic dependencies among point-cloud instances, thereby enabling cross-modal semantic enhancement and enriching point-cloud representations.
- Unlocking Instance Semantic Awareness for Domain Adaptive Semantic Segmentation. Fan Li, Xuan Wang, Min Qi, Zhaoxiang Zhang, Chengming Xu, Yuelei Xu. ICME 2025
🎖 Honors and Awards
- 2023.04: 🏆 National Silver Award, The 8th China International College Students’ “Internet+” Innovation and Entrepreneurship Competition
- 2022.08: 🏆 Gold Award (×2), The 8th China International “Internet+” Innovation and Entrepreneurship Competition (Shaanxi Provincial Level)
- 2022.07: 🏆 First Prize, 17th China Graduate Electronics Design Competition – Special Track (Preliminary Round)
- 2023.10: 🏆 Outstanding Graduate Student, Northwestern Polytechnical University
📖 Educations
- 2020.09 – Present, Ph.D. Program in Intelligent Unmanned Systems Science and Technology, School of Artificial Intelligence (formerly Institute of Unmanned Systems Technology), Northwestern Polytechnical University
- 2016.09 – 2020.06, B.E. in Computer Science and Technology, China University of Geosciences