| CARVIEW |
Mingyang Xie 谢铭阳
I am currently a 5th year Ph.D. student at the University of Maryland, advised by Christopher Metzler. Previously, I obtained my Bachelor degree from Washington University in St. Louis, advised by Ulugbek Kamilov.
I am broadly interested in computer vision and generative AI, with a focus on vision generation, multimodal understanding, and computational photography. I am actively looking for research internships starting Spring 2026.
Publications (* denotes equal contribution)
LaVR: Scene Latent Conditioned Generative Video Trajectory Re-Rendering using Large 4D Reconstruction Models
A camera-controlled video diffusion model that performs novel video trajectory synthesis by conditioning on 4D scene latents from a large 4D reconstruction model.
Can Hallucination Correction Improve Video-Language Alignment?
Aligns video and textual representations for spatiotemporal reasoning by letting the model learn to correct hallucinations.
Flash-Split: 2D Reflection Removal with Flash Cues and Latent Diffusion Separation
A diffusion-model-based approach for 2D reflection removal with latent-space separation.
Flash-Splat: 3D Reflection Removal with Flash Cues and Gaussian Splats
A simple yet effective approach for separating transmitted and reflected 3D scenes by using Gaussian Splatting and unpaired flash and no-flash multi-view images.
WaveMo: Learning Wavefront Modulations to See Through Scattering
Use a proxy reconstruction network to learn an optimal set of wavefront modulation patterns in an end-to-end fashion.
Snapshot High-Dynamic-Range Imaging with a Polarization Camera
A novel single-shot HDR imaging methodology using a polarization camera, achieving 4dB improvement over software-only baselines.
NeuWS: Neural Wavefront Shaping for Guidestar-Free Imaging Through Static and Dynamic Scattering Media
Neural signal representations enable breakthroughs in correcting for severe time-varying wavefront aberrations caused by scattering media.
TurbuGAN: An Adversarial Learning Approach to Spatially-Varying Multiframe Blind Deconvolution with Applications to Imaging Through Turbulence
Training GAN only on blurry images from a single scene to recover a sharp image without estimating the blur kernels or acquiring a large labelled dataset.
CoIL: Coordinate-Based Internal Learning for Tomographic Imaging
Learning the computed tomography measurement field by mapping X-ray angles to corresponding pixel values using implicit neural representation.
Joint Reconstruction and Calibration Using Regularization by Denoising with Application to Computed Tomography
A regularization by denoising approach for image reconstruction tasks where there exist parametric uncertainties in the imaging forward model.
PROVES: Establishing Image Provenance using Semantic Signatures
Using semantic signing for image verification against deep fake attacks.