Open-world Vision
Foundation models for open-vocabulary segmentation, detection, tracking, and compositional visual understanding.
Computer Vision · Embodied AI · Robotics
Feng Yan 颜峰
Ph.D. candidate in embodied intelligence. Research interests include multimodal models, vision-language-action systems, world models, robotic manipulation, autonomous systems, and visual tracking.
Research Interests
Generalizable perception and action for embodied intelligence.
Embodied AI
Vision-language-action models, world models, and Sim2Real learning for robotic manipulation and embodied decision making.
Autonomous Systems
Multimodal models for perception, prediction, planning, and generalization in autonomous driving and navigation.
Efficient AI Systems
Data-centric training, model compression, edge inference, and evaluation protocols for deployable embodied intelligence.
Open Source
Widely used projects in vision and embodied AI.
Selected repositories spanning open-world vision, DETR-based recognition, and Vision-Language-Action systems.
Core developer and contributor.
Grounded-Segment-Anything
Grounded SAM: open-world detect, segment, and generate by combining Grounding DINO with Segment Anything.
detrex
Research platform for DETR-based detection, segmentation, pose estimation, and related visual recognition tasks.
StarVLA
Lego-like codebase for Vision-Language-Action model development, training, and deployment.
Experience & Education
Appointments and education.
Embodied AI and multimodal perception
Research spans embodied AI, computer vision, and multimodal systems, with experience at enacta ai, Meituan, and ArcSoft.
Generalization, efficiency, and real-world grounding
Current work studies robotic manipulation, VLA models, world models, multi-camera perception, simulation, and evaluation.
Xi'an Jiaotong University
Ph.D. candidate in embodied intelligence; M.S., Xi'an Jiaotong University.
University of Electronic Science and Technology of China
B.S. in Automation, UESTC.
Selected Publications
Publications
Selected publications spanning embodied AI, robotic manipulation, multimodal models, tracking, and autonomous driving.
Contact
Contact
For research collaboration and academic inquiries, please contact by email or via Google Scholar.
Guestbook
Comments
Leave a short note. Spam is filtered; please keep it friendly.