Dongyoon Han

Affiliations

KAISTNAVER AI Lab

papers

1,704

total citations

papers (21)

Rethinking Spatial Dimensions of Vision Transformers

ICCV 2021arXiv

700

citations

OCR-Free Document Understanding Transformer

ECCV 2022arXiv

397

citations

Re-Labeling ImageNet: From Single to Multi-Labels, From Global to Localized Labels

CVPR 2021arXiv

164

citations

The Devil Is in the Points: Weakly Semi-Supervised Instance Segmentation via Point-Guided Mask Representation

CVPR 2023arXiv

citations

Demystifying the Neural Tangent Kernel From a Practical Perspective: Can It Be Trusted for Neural Architecture Search Without Training?

CVPR 2022arXiv

citations

Contrastive Vicinal Space for Unsupervised Domain Adaptation

ECCV 2022arXiv

citations

DaWin: Training-free Dynamic Weight Interpolation for Robust Adaptation

ICLR 2025arXiv

citations

NegMerge: Sign-Consensual Weight Merging for Machine Unlearning

ICML 2025arXiv

citations

Neglected Free Lunch - Learning Image Classifiers Using Annotation Byproducts

ICCV 2023arXiv

citations

Gramian Attention Heads are Strong yet Efficient Vision Learners

ICCV 2023arXiv

citations

Masking meets Supervision: A Strong Learning Alliance

CVPR 2025arXiv

citations

Morphing Tokens Draw Strong Masked Image Models

ICLR 2025arXiv

citations

Learning with Unmasked Tokens Drives Stronger Vision Learners

ECCV 2024arXiv

citations

SeiT++: Masked Token Modeling Improves Storage-efficient Training

ECCV 2024arXiv

citations

Token Bottleneck: One Token to Remember Dynamics

NEURIPS 2025arXiv

citations

Generating Instance-level Prompts for Rehearsal-free Continual Learning

ICCV 2023

citations

Dongyoon Han

Affiliations

papers (21)

Rethinking Spatial Dimensions of Vision Transformers

OCR-Free Document Understanding Transformer

Re-Labeling ImageNet: From Single to Multi-Labels, From Global to Localized Labels

Rethinking Channel Dimensions for Efficient Model Design

Model Stock: All we need is just a few fine-tuned models

Switching Temporary Teachers for Semi-Supervised Semantic Segmentation

DenseNets Reloaded: Paradigm Shift Beyond ResNets and ViTs

Scratching Visual Transformer's Back with Uniform Attention

The Devil Is in the Points: Weakly Semi-Supervised Instance Segmentation via Point-Guided Mask Representation

Demystifying the Neural Tangent Kernel From a Practical Perspective: Can It Be Trusted for Neural Architecture Search Without Training?

Contrastive Vicinal Space for Unsupervised Domain Adaptation

DaWin: Training-free Dynamic Weight Interpolation for Robust Adaptation

NegMerge: Sign-Consensual Weight Merging for Machine Unlearning

Neglected Free Lunch - Learning Image Classifiers Using Annotation Byproducts

Gramian Attention Heads are Strong yet Efficient Vision Learners

Masking meets Supervision: A Strong Learning Alliance

Morphing Tokens Draw Strong Masked Image Models

Learning with Unmasked Tokens Drives Stronger Vision Learners

SeiT++: Masked Token Modeling Improves Storage-efficient Training

Token Bottleneck: One Token to Remember Dynamics

Generating Instance-level Prompts for Rehearsal-free Continual Learning

papers (21)

Rethinking Spatial Dimensions of Vision Transformers

OCR-Free Document Understanding Transformer

Re-Labeling ImageNet: From Single to Multi-Labels, From Global to Localized Labels

Rethinking Channel Dimensions for Efficient Model Design

Model Stock: All we need is just a few fine-tuned models

Switching Temporary Teachers for Semi-Supervised Semantic Segmentation

DenseNets Reloaded: Paradigm Shift Beyond ResNets and ViTs

Scratching Visual Transformer's Back with Uniform Attention

The Devil Is in the Points: Weakly Semi-Supervised Instance Segmentation via Point-Guided Mask Representation

Demystifying the Neural Tangent Kernel From a Practical Perspective: Can It Be Trusted for Neural Architecture Search Without Training?

Contrastive Vicinal Space for Unsupervised Domain Adaptation

DaWin: Training-free Dynamic Weight Interpolation for Robust Adaptation

NegMerge: Sign-Consensual Weight Merging for Machine Unlearning

Neglected Free Lunch - Learning Image Classifiers Using Annotation Byproducts

Gramian Attention Heads are Strong yet Efficient Vision Learners

Masking meets Supervision: A Strong Learning Alliance

Morphing Tokens Draw Strong Masked Image Models

Learning with Unmasked Tokens Drives Stronger Vision Learners

SeiT++: Masked Token Modeling Improves Storage-efficient Training

Token Bottleneck: One Token to Remember Dynamics

Generating Instance-level Prompts for Rehearsal-free Continual Learning