[02/2026] Two paper accepted by ICRA 2026: "SURE: Semi-dense Uncertainty-REfined Feature Matching" and "Consistency-Driven Confidence Estimation for Stereo Matching"
[01/2026] One paper accepted by ICLR 2026: "Noise-Adaptive Diffusion Sampling for Inverse Problems Without Task-Specific Tuning".[Project Page]
[09/2025] Happy to be named among as "Top 2% Most Cited Scientists for 2023,2024,2025" by Stanford University.[Link]
[03/2025] Two paper accepted by MICCAI 2025: "Intra- and Cross-View Enhancement for OCTA Imaging" and "Improving OCTA Imaging through Cross-Domain Adaptation".
[04/2025] One paper accepted by ICLR 2025: "Uncertainty Aware Interest Point Detection and Description".
[11/2024] One paper accepted by TMLR: "When low-vision task meets dense prediction tasks with less data: an auxiliary self-trained geometry regularization".
[07/2024] One paper accepted by CVPR 2024: "Learning Intra-view and Cross-view Geometric Knowledge for Stereo Matching".
Featured Projects:
Inverse Problems (2024 ~ )
This research project focus on learning-based solutions for inverse problems,
including diffusion-based Bayesian sampling frameworks for robust image reconstruction and deep learning methods for enhanced OCT angiography imaging.
Z. Gu, J. Cheng, S. Leopold, B. Tan,
"A System and a Method of Generating Optical Coherence Tomography Angiography (OCTA) Images with Improved Image Quality",
Patent,
Y. Xia, Set. T., L. Zhen, Z. Gu,
"Noise-Adaptive Diffusion Sampling for Inverse Problems Without Task-Specific Tuning",
ICLR 2026[Project Page]
Trustworthy AI with Evidential Learning (2023 ~ )
This project investigates trustworthy AI through evidential learning for visual perception.
We develop uncertainty-aware models that estimate reliable confidence for tasks such as dense correspondence, stereo matching, and interest point detection, improving robustness under corruptions and challenging real-world conditions.
S. Lu, Z. Gu*,X. Jiang, J. Cheng,
"Consistency-Driven Confidence Estimation for Stereo Matching",
ICRA 2026,
"SURE: Semi-Dense Uncertainty-REfined Feature Matching",
ICRA 2026,
"Uncertainty Aware Interest Point Detection and Description",
WACV 2025,
Geometry-aware Regularzation for Dense Prediction (2022 ~ 2024)
This project focus on efficient geometry-aware regularization strategies for dense prediction tasks such as depth estimation and semantic segmentation.
The goal is to incorporate geometric priors into deep learning models to improve prediction consistency and structural coherence while maintaining computational efficiency.
Z. Gu,W. Liu, X. Yang, CS. Foo, J. Cheng,
"When low-vision task meets dense prediction tasks with less data: an auxiliary self-trained geometry regularization",
TMLR 2024,
"Learning intra-view and cross-view geometric knowledge for stereo matching",
CVPR 2024,
"SRNet: Self-supervised structure regularization for stereo matching",
Neurocomputing 2025,
Robotics Vision Perception (2020 ~ 2022):
My responsibility for this project is to apply the object detection and recognition algorithms to the company’s various types of robots, such as low-power robot toys for the customers and high-power industrial robots.
To be specific, we first adopt some advanced AI algorithms to locate and recognize hundreds of objects.
Then, we focus on the corner cases, where the existing AI algorithms fail on our actual scenes.
Finally, based on the deficiency of current methods, we modify and propose a new algorithm to enhance object recognition.
Now, our intelligent algorithm could recognize up to 300 categories of objects
The Outstanding Contribution Award, in Ubtech Research!
Intelligent Diagnosis of Fundus Diseases (2017 ~ 2020)
This project develops deep learning methods for automated diagnosis of fundus diseases such as diabetic retinopathy and glaucoma.
A key contribution is CE-Net, a context encoder network for medical image segmentation published in IEEE Transactions on Medical Imaging (TMI),
which has become an ESI Highly Cited Paper with over 2500 citations. Related work further explores noise-adaptive learning, vessel detection, and structure-preserving retinal image analysis for robust medical image understanding.
Z. Gu, J. Cheng, H. Fu, K. Zhou, H. Hao, Y. Zhao, T. Zhang, S. Gao and J. Liu,
"CE-Net: Context Encoder Network for 2D Medical Image Segmentation",
TMI 2019,[arXiv][Code]"ESI Highly Cited Paper"
"Noise adaptation generative adversarial network for medical image analysis",
TMI 2019,[Paper]
"Dense Dilated Network with Probability Regularized Walk for Vessel Detection",
TMI 2019,