Weijie Gan η”˜δΌŸζ·

I am a 5th year PhD student in Computational Imaging Group (CIG) from Computer Science and Engineering Department at Washington University in St. Louis (WashU), supervised by Prof. Ulugbek Kamilov and Prof. Hongyu An. Before starting my PhD, I received the M.Sc. degree in Computer Science, in 2020, also from WashU, and the B.Sc. degree in Automation from South China University of Technology, Guangzhou, China, in 2018.

My research interests are computational imaging, machine learning, computer vision, inverse problems, optimization and medical imaging.

My research focuses on algorithms designs and theoretical analysis for deep learning-based computational imaging. My work lies at the intersection of computational imaging, computer vision, machine learning, optimization, image processing, and physics. I am passionate about incorporating cross-domain knowledge to develop the state-of-the-art computer vision and generative AI models to solve the real-world challenging problems for various imaging systems as well as establishing theoretical analysis. My research topics include image reconstruction/restoration, inverse problems, diffusion models, image registration, self-supervised learning and correction of physical model uncertainty.

Email  /  CV  /  Linkedin  /  Twitter  /  Google Scholar

profile photo
News

2024/03: Our paper SPICER: Self-Supervised Learning for MRI with Automatic Coil Sensitivity Estimation and Reconstruction has been accepted to MRM.
2024/02: Our paper PtychoDV: Vision Transformer-Based Deep Unrolling Network for Ptychographic Image Reconstruction has been accepted to IEEE OJSP.
2024/01: Our paper A Plug-and-PlayImage Registration Network has been accepted to ICLR 2024.
2023/11: Two new preprints: Dose-aware Diffusion Model for 3D Ultra Low-dose PET Imaging and A Structured Pruning Algorithm for Model-based Deep Learning.
2023/10: New preprint: PtychoDV: Vision Transformer-Based Deep Unrolling Network for Ptychographic Image Reconstruction.
2023/10: New preprint: A Plug-and-PlayImage Registration Network.
2023/09: Our BC-PnP paper got accepted to NeurIPS 2023.
2023/07: Our SelfDEQ paper got accepted to IEEE TCI.
2023/05: Excited to share that I will join Siemens Healthineers, Knoxville, TN, as a summer research intern in 2023, supervised by Dr. Jorge Cabello and Dr. Maurizio Conti.
2022/05: Excited to share that I will join Los Alamos National Lab (LANL), Los Alamos, NM, as a summer research intern in 2022, supervised by Dr. Brendt Wohlberg.

Selected Research

My complete publications are available at google scholar. Representative papers are highlighted. (* denotes co-first authors)

Dose-aware Diffusion Model for 3D Ultra Low-dose PET Imaging
Huidong Xie*, Weijie Gan*, Bo Zhou, Xiongchao Chen, Qiong Liu, Xueqi Guo, Liang Guo, Hong An, Ulugbek Kamilov, Ge Wang, Chi Liu
arXiv, 2023
arXiv

A SOTA diffusion model for 3D ultra low dose whole-body PET denoising.

Efficient Model-based Deep Learning via Network Pruning and Fine-Tuning
Chicago Park*, Weijie Gan*, Zihao Zou, Yuyang Hu, Zhixin Sun, Ulugbek Kamilov
arXiv, 2023
arXiv

A novel application of network pruning for efficient deep model-based networks.

PtychoDV: Vision Transformer-Based Deep Unrolling Network for Ptychographic Image Reconstruction
Weijie Gan, Qiuchen Zhai, Michael Thompson McCann, Cristina Garcia-Cardona, Ulugbek Kamilov, Brendt Wohlberg
IEEE Open Journal of Signal Processing, 2024
arXiv

We proposed a new ptychographic image reconstruction algorithm based on vision transformer and deep unfolding.

A Plug-and-Play Image Registration Network
Junhao Hu*, Weijie Gan*, Zhixin Sun, Hong An, Ulugbek Kamilov
ICLR, 2024
arXiv / project page / OpenReview

The first plug-and-play method for image registration that uses a pre-trained denoiser as the prior information.

SPICER: Self-Supervised Learning for MRI with Automatic Coil Sensitivity Estimation and Reconstruction
Yuyang Hu*, Weijie Gan*, Chunwei Ying, Tongyao Wang, Cihat Eldeniz, Jiaming Liu, Yasheng Chen, Hong An, Ulugbek Kamilov
Magnetic Resonance in Medicine (MRM), 2024
arXiv / project page

The first self-supervised method for automatic coil sensitivity calibration in MRI.

Block Coordinate Plug-and-Play Methods for Blind Inverse Problems
Weijie Gan, Shirin Shoushtari, Yuyang Hu, Jiaming Liu, Hong An, Ulugbek Kamilov
NeurIPS 2023
arXiv / NeurIPS / project page

A block-coordinate PnP method with theoretical analysis for blind inverse problems.

Self-Supervised Deep Equilibrium Models With Theoretical Guarantees and Applications to MRI Reconstruction
Weijie Gan, Chunwei Ying, Parna Eshraghi, Tongyao Wang, Cihat Eldeniz, Yuyang Hu, Jiaming Liu, Yasheng Chen, Hong An, Ulugbek Kamilov
IEEE Transactions on Computational Imaging, 2023
arXiv / IEEE

First to propose self-supervised learning for deep equilibrium model with theoretical guarantees.

DeCoLearn: Deformation-compensated learning for image reconstruction without ground truth
Weijie Gan, Yu Sun, Cihat Eldeniz, Jiaming Liu, Hong An, Ulugbek Kamilov
IEEE Transactions on Medical Imaging, 2022
arXiv / project page

Joint image reconstruction and image registration without any ground-truth supervision.

CURE: Learning Cross-Video Neural Representations for High-Quality Frame Interpolation
Wentao Shangguan, Yu Sun, Weijie Gan, Ulugbek Kamilov
ECCV, 2022
arXiv / project page

First to propose a video frame interpolation algorithm based on neural field that can achieve SOTA performance.

Phase2Phase: Respiratory motion-resolved reconstruction of free-breathing magnetic resonance imaging using deep learning without a ground truth for improved liver imaging
Cihat Eldeniz*, Weijie Gan*, Sihao Chen, Tyler J Fraum, Daniel R Ludwig, Yan Yan, Jiaming Liu, Thomas Vahle, Uday Krishnamurthy, Ulugbek Kamilov, Hong An
Investigative Radiology, 2021
paper / media converge

Propose a new deep learning method to boost MRI results without requiring clean training data.

RARE: Image reconstruction using deep priors learned without groundtruth
Jiaming Liu, Yu Sun, Cihat Eldeniz, Weijie Gan, Hong An, Ulugbek Kamilov
IEEE Journal of Selected Topics in Signal Processing, 2020
arXiv

Exploit statistical priors specified by artifact-removing CNNs trained without groundtruth.

Interesting Projects
Track Microtubule in Medical Scan Video
CSE554 Geometric Computing for Biomedicine , 2018Fall
GitHub

Proposed method with practical GUI tool to detect and track single micro-tubule movement in scan video. The tools also record the growth of target microtubule during video.

The 11st National "NXP Cup" Smart Car Race Competition for College Students
Second Prize of South China University of Technology Division & Third Prize of the Southern China Division , May.2015 – Aug. 2016

Designed a an auto-run smart car, which could recognize a pure white road with black borders by a specifics cam-era(Linear CCD) captured only one-row data(single dimension).


Template is from Here.