research
Representation Learning-based Semantic Segmentation
[CVPR'24], [arXiv'24]
Unsupervised semantic segmentation
Weakly supervised semantic segmentation
Large Language Model for Medical Image Understanding
[arXiv'24]
Model-user communication of CXR Image through Multi-modal LLM
Generating medical report based on input medical image
Multi-modal Medical Image Synthesis
[arXiv'24]
Text-to-medical image generation with reinforcement learning
CT-to-MRI translation
fMRI-to-Image translation
PAC-Bayesian Domain Adaption Bounds
[UAI'22 - Best Paper Award], [Findings'22]
PAC-Bayesian domain adaptation bounds for multiclass learners
Domain adaptation bound for domain shift in discourse parsing
Allows a non-uniform sample complexity with no sampling
Large-scale empirical analyses on CV and NLP datasets
Domain Generalization/Adaptation
[UAI'22], [IJCAI'22], [Findings'22], [arXiv'21], [ISBI'21]
Domain generalization on CV, NLP, and medical imaging problems
Brain Lesion Segmentation
[ISBI'22a], [ISBI'22b]
White matter hyperintensity (WMH) segmentation
Brain lesion detection under domain shift in multi-site neuroimaging datasets
Multi-site Neuroimaging Data Harmonization
[biorxiv'22], [NeuroImage'21], [ICCV-W'21], [arXiv'21], [ISBI'21]
Harmonize multi-site/scanner neuroimaging data
PAC-Bayesian Performance Guarantee
[MICCAI'21]
Non-vacuous performance guarantee on skin cancer detection
Large-scale Video Analysis
[KDD'19]
Large-scale video annotation on YouTube-8M