Projects

Act to See: Structured Articulated Representations through Robot Interaction
A closed-loop robotic system that actively interacts with an unknown articulated object and reconstructs an explicit URDF online.
Jul 1, 2026
Code Project Paper
Act to See: Structured Articulated Representations through Robot Interaction
Winner of the CVPR 2026 OpenSUN3D Challenge
A winning submission, developed with Chengkun (Charlie) Li, for the Open-Vocabulary 3D Affordance Grounding track of the CVPR 2026 OpenSUN3D Challenge.
Jun 1, 2026
Project Winner Slides
Winner of the CVPR 2026 OpenSUN3D Challenge
Image-Pretrained Vision Transformers for Real-Time Traffic Anomaly Detection
This project explores the trade-offs of using image-pretrained ViTs for traffic anomaly detection, achieving 78.1% AUC-ROC at 200 FPS on the DoTA benchmark.
Feb 6, 2026
Image-Pretrained Vision Transformers for Real-Time Traffic Anomaly Detection
VRU tracking in LiDAR point cloud
Inspired by Spatio-Temporal Tracking paradigms, we leverage LiDAR data to track pedestrians in 3D space.
Oct 1, 2023
VRU tracking in LiDAR point cloud
YourArXiv.com -- a Personalized ArXiv with Embedding based recommendation
Inspired by language model for recommendation, we leverage Ada2 model to build a personalized ArXiv recommendation system.
Apr 20, 2023
YourArXiv.com -- a Personalized ArXiv with Embedding based recommendation
Deployment and optimization pipeline for YOLOv5 models on detection and segmentation tasks
Deployment and optimization pipeline for YOLOv5 models on detection and segmentation tasks
Dec 18, 2022 Object Detection
Deployment and optimization pipeline for YOLOv5 models on detection and segmentation tasks