| CARVIEW |
Jingxi Chen
I am a PhD student in the Computer Science Department at the University of Maryland, College Park. I am working with Prof. Yiannis Aloimonos and Cornelia Fermüller at Perception and Robotics Group. I also work closely with Prof. Christopher Metzler. During my master’s study I worked with Prof. Pratap Tokekar on using Reinforcement Learning in Multi-agent System research.
My research focuses on Video/Image Generative Models, neural representations for motion in videos, and 3D vision for robotics.
Selected Publications (* denotes equal contribution)
First Frame Is the Place to Go for Video Content Customization
In this work, we uncover a new perspective on video generation: the first frame acts as a conceptual memory buffer. Leveraging this insight, we achieve robust and generalized video content customization with just 20–50 training examples.
Repurposing Pre-trained Video Diffusion Models for Event-based Video Interpolation
In this work, we adapt pre-trained video diffusion models trained on internet-scale datasets to solve the specialized real-world video task of event-based video interpolation.
Temporally Consistent Atmospheric Turbulence Mitigation with Neural Representations
ConVRT is an efficient INR framework for video-based turbulence mitigation that operates in test-time optimization manner
Active Human Pose Estimation via an Autonomous UAV Agent
We leverage radiance fields to imagine different human views to find the best drone pose for aerial cinematography.
Microsaccade-inspired Event Camera for Robotics
Inspired by microsaccades, we designed an event-based perception system capable of simultaneously maintaining low reaction time and stable texture.
CodedEvents: Optimal Point-Spread-Function Engineering for 3D-Tracking with Event Cameras
CodedEvents is a novel method for optimal point-spread-function engineering for 3D-tracking with event cameras.
Proxmap: Proximal occupancy map prediction for efficient indoor robot navigation
We present a self-supervised occupancy prediction technique, ProxMaP, to predict the occupancy within the proximity of the robot to enable faster navigation.
Multi-Agent Reinforcement Learning for Visibility-based Persistent Monitoring
We present a Multi-Agent Reinforcement Learning (MARL) algorithm for the Visibility-based Persistent Monitoring (VPM) problem.
Template from Keunhong Park