Dongyoon Han
E-mails: dongyoon.han at navercorp.com / karusun at gmail.com
I am a Senior Research Scientist in NAVER AI Lab (from 2018.01 - ) and an Adjunct Professor at KAIST GSAI (from 2021.09 - ).
I am passionate about designing cutting-edge deep neural networks and pioneering training strategies for large language models and vision-language models from a machine learning perspective.
Please reach out if you're excited about working together on these topics 🙏🙏
Research interests
Efficient/large-scale machine learning
Language/vision modeling
Representation learning
Multi-modal learning
Model editing/merging/unlearning
News
🆕 Three papers have been accepted at ICLR 2025
Two papers have been accepted at NeurIPS 2024 workshops
One paper has been accepted at NeurIPS 2024
One paper has been accepted at ACCV 2024!
Eight papers have been accepted at ECCV 2024
One paper has been accepted at NeurIPS 2023
Four papers have been accepted at ICCV 2023
Will serve as an AC for NeurIPS 2023 Datasets and Benchmarks
One paper has been accepted at IJCAI 2023
One paper has been accepted at CVPR 2023
One paper has been accepted at Neural Networks (IF=9.657)
Two papers have been accepted at AAAI 2023
Two papers have been accepted at ECCV 2022
One paper has been accepted at ICML 2022
One paper has been accepted at CVPR 2022
Two papers have been accepted at ICLR 2022
One paper has been accepted at ICCV 2021
One paper has been accepted at Pattern Recognition (IF =7.74)
Two papers have been accepted at CVPR 2021
One paper has been accepted at ICLR 2021
Academic activities
Served (or will serve) as an AC at the following conferences:
NeurIPS 2024 Datasets and Benchmarks
NeurIPS 2023 Datasets and Benchmarks
Served (or will serve) as a reviewer at the following conferences:
ICLR 2025, CVPR 2025, ICML 2025
AAAI 2024, ICLR 0224, CVPR 2024, ICML 2024, IJCAI 2024, ECCV 2024, NeurIPS 2024
ICLR 2023, CVPR2023, ICML 2023, ICCV 2023, NeurIPS 2023
ICLR 2022, CVPR 2022, ICML 2022, ECCV 2022, NeurIPS 2022
AAAI 2021, ICLR 2021, CVPR 2021 (outstanding reviewer), ICML 2021, ICCV 2021, NeurIPS 2021
AAAI 2020, ICLR 2020, CVPR 2020, ECCV 2020, NeurIPS 2020
ICLR 2019, ICML 2019, ICCV 2019, NeurPS 2019,
CVPR 2018, NeurIPS 2018 (top-200 reviewers)
Served as a reviewer for TMLR, TKDE, TNNLS, TPAMI, and TIP.
Delivered a class AI599 at KAIST
Awards
4th place at the last ImageNet competition on ILSVRC 2017 object localization task
Outstanding Paper Award at ICACT 2014
Publications
Masked Image Modeling via Dynamic Token Morphing
ICLR 2025
Taekyung Kim*, Byeongho Heo, Dongyoon Han* (* equal contribution)
[Paper]
DaWin: Training-free Dynamic Weight Interpolation for Robust Adaptation
ICLR 2025
Changdae Oh, Yixuan Li, Kyungwoo Song*, Sangdoo Yun*, Dongyoon Han* (* equal advising)
[Paper]
Tint Your Models Task-wise for Improved Multi-task Model Merging
arxiv 2024
Aecheon Jung, Seunghwan Lee, Dongyoon Han*, Sungeun Hong* (* equal advising)
Model Stock: All we need is just a few fine-tuned models
ECCV 2024
Dong-Hwan Jang, Sangdoo Yun*, Dongyoon Han* (* equal advising)
Thanks to our insights in the fine-tuned weight space, fine-tuning a few models (i.e., only two) can lead to superior merged weights (closer to the center of a weight space) without merging many fine-tuned models under extensive parameter searches like Model Soup.
DenseNets Reloaded: Paradigm Shift Beyond ResNets and ViTs
ECCV 2024
Donghyun Kim*, Byeongho Heo, Dongyoon Han* (* equal contribution)
Through strategic enhancements, we revitalize the once-superior DenseNets, surpassing milestones like ConvNeXts and MogaNets, and recent frontiers ViT-hybrids. By focusing on expanding transition layers and refining the architecture's building blocks, we've crafted SOTA architectures on ImageNet.
Rotary Position Embedding for Vision Transformer
ECCV 2024
Byeongho Heo, Song Park, Dongyoon Han, Sangdoo Yun
We show that vision transformers can be enhanced by utilizing RoPE, which allows them to extrapolate beyond the resolutions they were trained on very effectively. Furthermore, our variant RoPE-mixed, which uses learnable frequencies for both axes instead of alternating dimensions for each axis as in traditional 2D RoPE, performs better for 2D signals.
[Paper]
Learning with Unmasked Tokens Drives Stronger Vision Learners
ECCV 2024
Taekyung Kim*, Sanghyuk Chun, Byeongho Heo, Dongyoon Han*
(* equal contribution)
Our insight is that the limited attention span of a MIM pre-trained encoder is attributed to MIM's sole focus on regressing masked tokens only, which hampers the encoder's broader context learning. We, therefore, explicitly incorporate unmasked tokens into the training process, which enables the encoder to learn from broader context supervision with resulting expansive attention maps.
[Paper]
Similarity of Neural Architectures Based on Input Gradient Transferability
ECCV 2024
Jaehui Hwang, Dongyoon Han, Byeongho Heo, Song Park, Sanghyuk Chun, Jong-Seok Lee
We observe that adversarial attack transferability may reveal information about input gradients and decision boundaries reflecting similarities across models. Our large-scale analysis of 69 state-of-the-art ImageNet-pre-trained classifiers using our proposed similarity function SAT confirms this observation.
[Paper]
SeiT++: Masked Token Modeling Improves Storage-efficient Training
ECCV 2024
Minhyun Lee, Song Park, Byeongho Heo, Dongyoon Han, Hyunjung Shim
Built upon the recent breakthrough SeiT with Vector-Quantized (VQ) techniques, SeiT++ eliminates the need for training labels by integrating Masked Token Modeling (MTM) for self-supervised pre-training. SeiT++ significantly enhances SeiT and reaches an ImageNet top-1 accuracy of 77.8% using only 1% storage size of the full ImageNet-1K data.
NegMerge: Consensual Weight Negation for Strong Machine Unlearning
arxiv 2024
Hyoseo Kim, Dongyoon Han*, Junsuk Choe* (* equal advising)
Task-arithmetic-based merging suffers from the vulnerability of fine-tuned models during unlearning. This method addresses the issue by merging all the searched weights instead of relying on a single weight for unlearning.
Leveraging Temporal Contextualization for Video Action Recognition
ECCV 2024
Minji Kim, Dongyoon Han, Taekyung Kim, Bohyung Han
Video understanding could be enhanced by leveraging temporal information through global interactions via Temporal Contextualization (TC), a layer-wise temporal information infusion mechanism.
HYPE: Hyperbolic Entailment Filtering for Underspecified Images and Texts
ECCV 2024
Wonjae Kim, Sanghyuk Chun, Taekyung Kim, Dongyoon Han, Sangdoo Yun
Using hyperbolic embeddings and entailment cones, our HYPE could very effectively evaluate and filter out samples with meaningless or underspecified semantics, enhancing data specificity.
[Paper]
Match me if you can: Semantic Correspondence Learning with Unpaired Images
ACCV 2024
Jiwon Kim, Byeongho Heo, Sangdoo Yun, Seungryong Kim, Dongyoon Han*
(* corresponding author)
[Paper]
Neglected Free Lunch - Learning Image Classifiers Using Annotation Byproducts
ICCV 2023
Dongyoon Han*, Junsuk Choe*, Seonghyeok Chun, John Joon Young Chung, Minsuk Chang, Sangdoo Yun, Jean Y. Song, Seong Joon Oh
(* equal contribution)
Switching Temporary Teachers for Semi-Supervised Semantic Segmentation
NeurIPS 2023
Jaemin Na, Jung-Woo Ha, Hyung Jin Chang, Dongyoon Han*, Wonjun Hwang*
(* equal advising)
Augmenting Sub-model to Improve Main Model
arxiv 2023
Byeongho Heo, Taekyung Kim, Sangdoo Yun, Dongyoon Han
Generating Instance-level Prompts for Rehearsal-free Continual Learning
ICCV 2023 (oral presentation)
Dahuin Jung, Dongyoon Han, Jiwhan Bang, Hwanjun Song
Gramian Attention Heads are Strong yet Efficient Vision Learners
ICCV 2023
Jongbin Ryu*, Dongyoon Han*, Jongwoo Lim
(* equal contribution)
[Paper][Project Page]
Scratching Visual Transformer's Back with Uniform Attention
ICCV 2023
Nam Hyeon-Woo, Kim Yu-Ji, Byeongho Heo, Dongyoon Han, Seong Joon Oh, Tae-Hyun Oh
GeNAS: Neural Architecture Search with Better Generalization
IJCAI 2023
Joonhyun Jeong, Joonsang Yu, Geondo Park, Dongyoon Han, YoungJoon Yoo
The Devil is in the Points: Weakly Semi-Supervised Instance Segmentation via Point-Guided Mask Representation
CVPR 2023
Beomyoung Kim, Joonhyun Jeong, Dongyoon Han, Sung Ju Hwang
TL-ADA: Transferable Loss-based Active Domain Adaptation
Neural Networks 2023
Kyeongtak Han, Youngeun Kim, Dongyoon Han, Sungeun Hong
Can We Find Strong Lottery Tickets in Generative Models?
AAAI 2023
Sangyeop Yeo, Yoojin Jang, Jy-yong Sohn, Dongyoon Han, Jaejun Yoo
Spatiotemporal Augmentation on Selective Frequencies for Video Representation Learning
AAAI 2023
Jinhyung Kim, Taeoh Kim, Minho Shim, Dongyoon Han*, Dongyoon Wee, Junmo Kim
AAAI 2023
Contrastive Vicinal Space for Unsupervised Domain Adaptation
ECCV 2022
Jaemin Na, Dongyoon Han, Hyung Jin Chang, and Wonjun Hwang
Donut: Document Understanding Transformer without OCR
ECCV 2022
Geewook Kim, Teakgyu Hong, Moonbin Yim, Jinyoung Park, Jinyeong Yim, Wonseok Hwang, Sangdoo Yun, Dongyoon Han, and Seunghyun Park
Time Is MattEr: Temporal Self-supervision for Video Transformers
ICML 2022
Sukmin Yun, Jaehyung Kim, Dongyoon Han, Hwanjun Song, Jung-Woo Ha, Jinwoo Shin
Neural Architecture Search with Loss Flatness-aware Measure
ICML 2022, Workshop on Dynamic Neural Networks
Joonhyun Jeong, Joonsang Yu, Dongyoon Han, YoungJoon Yoo
An Extendable, Efficient and Effective Transformer-based Object Detector
arxiv, under review
Hwanjun Song, Deqing Sun, Sanghyuk Chun, Varun Jampani, Dongyoon Han, Byeongho Heo, Wonjae Kim, and Ming-Hsuan Yang
Demystifying the Neural Tangent Kernel from a Practical Perspective: Can it be trusted for Neural Architecture Search without training?
CVPR 2022
Jisoo Mok, Byunggook Na, Ji-Hoon Kim*, Dongyoon Han*, Sungroh Yoon*
(* equal advising)
Learning Features with Parameter-Free Layers
ICLR 2022, Best paper of NAVER AI Lab 2022
Dongyoon Han, YoungJoon Yoo, Beomyoung Kim, and Byeongho Heo
ViDT: An Efficient and Effective Fully Transformer-based Object Detector
ICLR 2022
Hwanjun Song, Deqing Sun, Sanghyuk Chun, Varun Jampani, Dongyoon Han, Byeongho Heo, Wonjae Kim, and Ming-Hsuan Yang
Rethinking spatial dimensions of vision transformers
ICCV 2021
Byeongho Heo, Sangdoo Yun, Dongyoon Han, Sanghyuk Chun, Junsuk Choe, and Seong Joon Oh
Detecting and Removing Text in the Wild
IEEE Access, vol. 9, pp. 123313-123323, 2021
Junho Cho, Sangdoo Yun, Dongyoon Han, Byeogho Heo, and Jin Young Choi
Region-based dropout with attention prior for weakly supervised object localization
Pattern Recognition, 116, 2021
Junsuk Choe, Dongyoon Han, Sangdoo Yun, Jung-Woo Ha, Seong Joon Oh, and Hyunjung Shim
Rethinking Channel Dimensions for Efficient Model Design
CVPR 2021
Dongyoon Han, Sangdoo Yun, Byeongho Heo, and YoungJoon Yoo
Re-labeling ImageNet: from Single to Multi-Labels, from Global to Localized Labels
CVPR 2021
Sangdoo Yun, Seong Joon Oh, Byeongho Heo, Dongyoon Han, Junsuk Choe, and Sanghyuk Chun
Slowing Down the Weight Norm Increase in Momentum-based Optimizers
ICLR 2021
Byeongho Heo, Sanghyuk Chun, Seong Joon Oh, Dongyoon Han, Sangdoo Yun, Youngjung Uh, and Jung-Woo Ha
VideoMix: Rethinking Data Augmentation for Video Classification
arxiv, under review
Sangdoo Yun, Seong Joon Oh, Byeongho Heo, Dongyoon Han, and Jinhyung Kim
An Empirical Evaluation on Robustness and Uncertainty of Regularization Methods
ICML 2019, Uncertainty & Robustness in Deep Learning Workshop
Sanghyuk Chun, Seong Joon Oh, Sangdoo Yun, Dongyoon Han, Junsuk Choe, YoungJoon Yoo
Unpaired Sketch-to-Line Translation via Synthesis of Sketches
SIGGRAPH Asia 2019, Technical Briefs
Gayoung Lee, Dohyun Kim, Youngjoon Yoo, Dongyoon Han, Jung-Woo Ha, Jaehyuk Chang
EXTD: Extremely Tiny Face Detector via Iterative Filter Reuse
arxiv, under review
Youngjoon Yoo, Dongyoon Han, and Sangdoo Yun
CutMix:Regularization Strategy to Train Strong Classifiers with Localizable Features
ICCV 2019 (oral presentation)
Sangdoo Yun, Dongyoon Han, Seong Joon Oh, Sanghyuk Chun, Junsuk Choe, and Youngjoon Yoo
What is wrong with scene text recognition model comparisons? dataset and model analysis
ICCV 2019 (oral presentation)
Jeonghun Baek, Geewook Kim, Junyeop Lee, Sungrae Park, Dongyoon Han, Sangdoo Yun, Seong Joon Oh, and Hwalsuk Lee
Where to be adversarial perturbations added? Investigating and manipulating pixel robustness using input gradients
ICLR 2019, Debugging Machine Learning Models Workshop
Jisung Hwang, Younghoon Kim, Sanghyuk Chun, Jaejun Yoo, Ji-Hoon Kim, and Dongyoon Han*
Character Region Awareness for Text Detection
CVPR 2019
Youngmin Baek, Bado Lee, Dongyoon Han, Sangdoo Yun, and Hwalsuk Lee,
Concentrated-Comprehensive Convolutions for lightweight semantic segmentation
arxiv, under review
Hyojin Park, Youngjoon Yoo, Geonseok Seo, and Dongyoon Han, Sangdoo Yun, Nojun Kwak
Learning Receptive Field Size by Learning Filter Size
WACV 2019
Yekang Lee, Heechul Jung, Dongyoon Han, Kyungsu Kim, and Junmo Kim
Towards Flatter Loss Surface via Nonmonotonic Learning Rate Scheduling
UAI 2018
Sihyeon Seong, Yekang Lee, Youngwook Kee, Dongyoon Han, and Junmo Kim
Unified Simultaneous Clustering and Feature Selection for Unlabeled and Labeled Data
IEEE Transactions on Neural Networks and Learning Systems (TNNLS), 2018
Dongyoon Han and Junmo Kim
Deep Pyramidal Residual Networks
CVPR 2017
Dongyoon Han, Jiwhan Kim, and Junmo Kim
Salient Region Detection via High-Dimensional Color Transform and Local Spatial Support
IEEE Transactions on Image Processing (TIP), 2016
Jiwhan Kim, Dongyoon Han, Yu-Wing Tai, and Junmo Kim
Unsupervised Orthogonal Basis Feature Selection
CVPR 2015
Dongyoon Han and Junmo Kim
Salient Region Detection via High-Dimensional Color Transform
Jiwhan Kim, Dongyoon Han, Yu-Wing Tai, and Junmo Kim
CVPR 2014
Publications
2024
Aecheon Jung, Seunghwan Lee, Dongyoon Han*, Sungeun Hong*, "Tint Your Models Task-wise for Improved Multi-task Model Merging", arXiv:2412.19098. [PDF] (* corresponding author)
Minhyun Lee, Seungho Lee, Song Park, Dongyoon Han, Byeongho Heo, Hyunjung Shim, "MaskRIS: Semantic Distortion-aware Data Augmentation for Referring Image Segmentation", arXiv:2411.19067. [PDF]
Hyoseo Kim, Dongyoon Han*, Junsuk Choe*, "NegMerge: Consensual Weight Negation for Strong Machine Unlearning", arXiv:2410.05583. [PDF] (* corresponding author)
Changdae Oh, Sharon Li, Kyungwoo Song*, Sangdoo Yun*, Dongyoon Han*, "DaWin: Training-free Dynamic Weight Interpolation for Robust Adaptation", arXiv:2410.03782. [PDF] (* corresponding author)
Jung Hyun Lee, June Yong Yang, Byeongho Heo, Dongyoon Han, Kang Min Yoo, "Token-Supervised Value Models for Enhancing Mathematical Reasoning Capabilities of Large Language Models", arXiv:2407.12863. [PDF]
Changdae Oh, Hyesu Lim, Mijoo Kim, Dongyoon Han, Sangdoo Yun, Jaegul Choo, Alexander Hauptmann, Zhi-Qi Cheng, Kyungwoo Song, "Towards Calibrated Robust Fine-Tuning of Vision-Language Models", Advances in Neural Information Processing Systems (NeurIPS), 2024. [PDF]
Jiwon Kim, Byeongho Heo, Sangdoo Yun, Seungryong Kim, Dongyoon Han*, "Match me if you can: Semantic Correspondence Learning with Unpaired Images", Asian Conference on Computer Vision (ACCV) 2024 . [PDF] (* corresponding author)
Taekyung Kim*, Sanghyuk Chun, Byeongho Heo, Dongyoon Han*, "Longer-range Contextualized Masked Autoencoder", European Conference on Computer Vision (ECCV), 2024. [PDF] (* 1st author)
Dong-Hwan Jang, Sangdoo Yun*, Dongyoon Han*, "Model Stock: All we need is just a few fine-tuned models", European Conference on Computer Vision (ECCV), 2024. [PDF] (* corresponding author)
Donghyun Kim*, Byeongho Heo, Dongyoon Han*, "DenseNets Reloaded: Paradigm Shift Beyond ResNets and ViTs", European Conference on Computer Vision (ECCV), 2024. [PDF] (* 1st author, DH initiated the project)
Byeongho Heo, Song Park, Dongyoon Han, Sangdoo Yun, "Rotary Position Embedding for Vision Transformer", European Conference on Computer Vision (ECCV), 2024. [PDF].
Wonjae Kim, Sanghyuk Chun, Taekyung Kim, Dongyoon Han, Sangdoo Yun, "HYPE: Hyperbolic Entailment Filtering for Underspecified Images and Texts", European Conference on Computer Vision (ECCV), 2024. [PDF]
Minji Kim, Dongyoon Han, Taekyung Kim, Bohyung Han, "Leveraging Temporal Contextualization for Video Action Recognition", European Conference on Computer Vision (ECCV), 2024. [PDF] (co-mentored)
Minhyun Lee*, Song Park*, Byeongho Heo, Dongyoon Han, Hyunjung Shim, "SeiT++: Masked Token Modeling Improves Storage-efficient Training", European Conference on Computer Vision (ECCV), 2024. [PDF] (co-mentored)
Jaehui Hwang, Dongyoon Han, Byeongho Heo, Song Park, Sanghyuk Chun, Jong-Seok Lee, "Similarity of Neural Architectures Based on Input Gradient Transferability Authors", European Conference on Computer Vision (ECCV), 2024. [PDF] (co-mentored)
Taekyung Kim*, Byeongho Heo, Dongyoon Han*, "Masked Image Modeling via Dynamic Token Morphing", arXiv:2401.00254, [PDF] (* 1st author)
2023
Jaemin Na, Jung-Woo Ha, Hyung Jin Chang, Dongyoon Han*, Wonjun Hwang*, "Switching Temporary Teachers for Semi-Supervised Semantic Segmentation", Advances in Neural Information Processing Systems (NeurIPS), 2023. [PDF] (* corresponding author)
Byeongho Heo, Taekyung Kim, Sangdoo Yun, Dongyoon Han, "Augmenting Sub-model to Improve Main Model", arXiv:2306.11339. [PDF]
Dongyoon Han*, Junsuk Choe*, Seonghyeok Chun, John Joon Young Chung, Minsuk Chang, Sangdoo Yun, Jean Y. Song, Seong Joon Oh, "Neglected Free Lunch -- Learning Image Classifiers Using Annotation Byproducts", IEEE International Conference on Computer Vision (ICCV), 2023. [PDF] (* 1st author)
Jongbin Ryu*, Dongyoon Han*, Jongwoo Lim, “Gramian Attention Heads are Strong yet Efficient Vision Learners”, IEEE International Conference on Computer Vision (ICCV), 2023. [PDF] (* 1st author)
Dahuin Jung, Dongyoon Han, Jiwhan Bang, Hwanjun Song, “Generating Instance-level Prompts for Rehearsal-free Continual Learning”, IEEE International Conference on Computer Vision (ICCV), 2023. [PDF] (co-mentored)
Nam Hyeon-Woo, Kim Yu-Ji, Byeongho Heo, Dongyoon Han, Seong Joon Oh, Tae-Hyun Oh, "Scratching Visual Transformer's Back with Uniform Attention N Hyeon-Woo", IEEE International Conference on Computer Vision (ICCV), 2023. [PDF] (co-mentored)
Beomyoung Kim, Joonhyun Jeong, Dongyoon Han, Sung Ju Hwang, "The Devil is in the Points: Weakly Semi-Supervised Instance Segmentation via Point-Guided Mask Representation", IEEE International Conference on Computer Vision and Pattern Recognition (CVPR), 2023. [PDF]
Kyeongtak Han, Youngeun Kim, Dongyoon Han, Sungeun Hong, "TL-ADA: Transferable Loss-based Active Domain Adaptation", Neural Networks, 2023. [PDF]
Sangyeop Yeo, Yoojin Jang, Jy-yong Sohn, Dongyoon Han, Jaejun Yoo, "Can We Find Strong Lottery Tickets in Generative Models?", In Proceedings of the AAAI Conference on Artificial Intelligence (AAAI) 2023. [PDF]
Jinhyung Kim, Taeoh Kim, Minho Shim, Dongyoon Han, Dongyoon Wee, Junmo Kim, "Spatiotemporal Augmentation on Selective Frequencies for Video Representation Learning", In Proceedings of the AAAI Conference on Artificial Intelligence (AAAI) 2023. [PDF] (co-mentored)
2022
Jaemin Na, Dongyoon Han, Hyung Jin Chang, and Wonjun Hwang, “Contrastive Vicinal Space for Unsupervised Domain Adaptation”, European Conference on Computer Vision (ECCV), 2022. [PDF]
Geewook Kim, Teakgyu Hong, Moonbin Yim, Jinyoung Park, Jinyeong Yim, Wonseok Hwang, Sangdoo Yun, Dongyoon Han, and Seunghyun Park, “Donut: Document Understanding Transformer without OCR”, European Conference on Computer Vision (ECCV), 2022. [PDF]
Sukmin Yun, Jaehyung Kim, Dongyoon Han, Hwanjun Song, Jung-Woo Ha, Jinwoo Shin, "Time Is MattEr: Temporal Self-supervision for Video Transformers", International Conference on Machine Learning (ICML) 2022. [PDF]
Joonhyun Jeong, Joonsang Yu, Dongyoon Han, YoungJoon Yoo, "Neural Architecture Search with Loss Flatness-aware Measure", International Conference on Machine Learning (ICML), 2022, Workshop on Dynamic Neural Networks. [PDF]
Kyeongtak Han, Youngeun Kim, Dongyoon Han, Sungeun Hong, "Loss-based Sequential Learning for Active Domain Adaptation", arXiv preprint arXiv:2204.11665. [PDF]
Hwanjun Song, Deqing Sun, Sanghyuk Chun, Varun Jampani, Dongyoon Han, Byeongho Heo, Wonjae Kim, and Ming-Hsuan Yang, "An Extendable, Efficient and Effective Transformer-based Object Detector", arXiv preprint arXiv:2204.07962. [PDF][CODE]
Jisoo Mok, Byunggook Na, Ji-Hoon Kim*, Dongyoon Han*, Sungroh Yoon*, "Demystifying the Neural Tangent Kernel from a Practical Perspective: Can it be trusted for Neural Architecture Search without training?" IEEE International Conference on Computer Vision and Pattern Recognition (CVPR), 2022. [PDF][CODE] (* corresponding author)
Dongyoon Han, YoungJoon Yoo, Beomyoung Kim, and Byeongho Heo, "Learning Features with Parameter-Free Layers", International Conference on Learning Representations (ICLR), 2022, Best paper among NAVER AI Lab's 2022 papers, [PDF][CODE][Openreview]
Hwanjun Song, Deqing Sun, Sanghyuk Chun, Varun Jampani, Dongyoon Han, Byeongho Heo, Wonjae Kim, and Ming-Hsuan Yang, "ViDT: An Efficient and Effective Fully Transformer-based Object Detector", International Conference on Learning Representations (ICLR), 2022. [PDF][CODE][Openreview]
2021
Byeongho Heo, Sangdoo Yun, Dongyoon Han, Sanghyuk Chun, Junsuk Choe, and Seong Joon Oh, "Rethinking spatial dimensions of vision transformers", IEEE International Conference on Computer Vision (ICCV), 2021. [PDF][CODE]
Junho Cho, Sangdoo Yun, Dongyoon Han, Byeogho Heo, and Jin Young Choi, "Detecting and Removing Text in the Wild," in IEEE Access, vol. 9, pp. 123313-123323, 2021. [PDF]
Junsuk Choe, Dongyoon Han, Sangdoo Yun, Jung-Woo Ha, Seong Joon Oh, and Hyunjung Shim, "Region-based dropout with attention prior for weakly supervised object localization", Pattern Recognition (impact factor: 7.74), 116, 2021.
Dongyoon Han, Sangdoo Yun, Byeongho Heo, and YoungJoon Yoo, "Rethinking Channel Dimensions for Efficient Model Design", IEEE International Conference on Computer Vision and Pattern Recognition (CVPR), 2021. [PDF][CODE]
Sangdoo Yun, Seong Joon Oh, Byeongho Heo, Dongyoon Han, Junsuk Choe, and Sanghyuk Chun, "Re-labeling ImageNet: from Single to Multi-Labels, from Global to Localized Labels", IEEE International Conference on Computer Vision and Pattern Recognition (CVPR), 2021. [PDF][CODE]
Byeongho Heo*, Sanghyuk Chun*, Seong Joon Oh, Dongyoon Han, Sangdoo Yun, Youngjung Uh, and Jung-Woo Ha, "Slowing Down the Weight Norm Increase in Momentum-based Optimizers", International Conference on Learning Representations (ICLR), 2021, [PDF][CODE][Openreview]
Before 2020
Sangdoo Yun, Seong Joon Oh, Byeongho Heo, Dongyoon Han, and Jinhyung Kim, "VideoMix: Rethinking Data Augmentation for Video Classification." arXiv preprint arXiv:2012.03457 (2020). [PDF]
Sanghyuk Chun, Seong Joon Oh, Sangdoo Yun, Dongyoon Han, Junsuk Choe, YoungJoon Yoo, "An Empirical Evaluation on Robustness and Uncertainty of Regularization methods", International Conference on Machine Learning (ICML), 2019, Uncertainty & Robustness in Deep Learning Workshop. [PDF]
Gayoung Lee, Dohyun Kim, Youngjoon Yoo, Dongyoon Han, Jung-Woo Ha, Jaehyuk Chang, "Unpaired Sketch-to-Line Translation via Synthesis of Sketches", SIGGRAPH Asia, 2019 Technical Briefs. [PDF]
Youngjoon Yoo, Dongyoon Han, and Sangdoo Yun, "EXTD: Extremely Tiny Face Detector via Iterative Filter Reuse", arXiv preprint arXiv:1906.06579. [PDF]
Sangdoo Yun, Dongyoon Han, Seong Joon Oh, Sanghyuk Chun, Junsuk Choe, and Youngjoon Yoo, “CutMix:Regularization Strategy to Train Strong Classifiers with Localizable Features”, IEEE International Conference on Computer Vision (ICCV), 2019 (oral). [PDF] [CODE]
Jeonghun Baek, Geewook Kim, Junyeop Lee, Sungrae Park, Dongyoon Han, Sangdoo Yun, Seong Joon Oh, and Hwalsuk Lee, “What is wrong with scene text recognition model comparisons? dataset and model analysis”, IEEE International Conference on Computer Vision (ICCV), 2019 (oral). [PDF] [CODE]
Jisung Hwang*, Younghoon Kim*, Sanghyuk Chun*, Jaejun Yoo, Ji-Hoon Kim, and Dongyoon Han†, “Where to be adversarial perturbations added? Investigating and manipulating pixel robustness using input gradients”, International Conference on Learning Representations (ICLR), 2019, Debugging Machine Learning Models Workshop. [PDF] († corresponding author)
Hyojin Park, Youngjoon Yoo, Geonseok Seo, and Dongyoon Han, Sangdoo Yun, Nojun Kwak, “Concentrated-Comprehensive Convolutions for lightweight semantic segmentation”, arXiv preprint arXiv:1812.04920. [PDF]
Youngmin Baek, Bado Lee, Dongyoon Han, Sangdoo Yun, and Hwalsuk Lee, “Character Region Awareness for Text Detection”, IEEE Computer Vision and Pattern Recognition (CVPR), 2019. [PDF] [CODE]
Yekang Lee, Heechul Jung, Dongyoon Han, Kyungsu Kim, and Junmo Kim, “Learning Receptive Field Size by Learning Filter Size”, IEEE Winter Conference on Applications of Computer Vision (WACV), 2019. [PDF]
Sihyeon Seong, Yekang Lee, Youngwook Kee, Dongyoon Han, and Junmo Kim, “Towards Flatter Loss Surface via Nonmonotonic Learning Rate Scheduling”, Conference on Uncertainty in Artificial Intelligence (UAI), 2018 (acceptance rate: 24.6%). [PDF]
Dongyoon Han and Junmo Kim, "Unified Simultaneous Clustering and Feature Selection for Unlabeled and Labeled Data," IEEE Transactions on Neural Networks and Learning Systems (TNNLS), 2018 (impact factor: 11.683). [PDF]
Dongyoon Han*, Jiwhan Kim*, and Junmo Kim, “Deep Pyramidal Residual Networks”, IEEE International Conference on Computer Vision and Pattern Recognition (CVPR), 2017. [PDF][CODE]
Juseung Yun, Jaeyoung Lee, Dongyoon Han, Jeongwoo Ju, and Junmo Kim, “Cost-efficient 3D face reconstruction from a single 2D image”, ICACT, 2017.
Jiwhan Kim, Dongyoon Han, Wonjun Hwang, and Junmo Kim, “3D face recognition via discriminative keypoint selection”, 14th International Conference on Ubiquitous Robots and Ambient Intelligence (URAI), 2017.
Jiwhan Kim, Dongyoon Han, Yu-Wing Tai, and Junmo Kim, "Salient Region Detection via High-Dimensional Color Transform and Local Spatial Support," IEEE Transactions on Image Processing (TIP), Vol. 25, No. 1, pp. 9-23, 2016 (impact factor: 6.79). [PDF]
Jiwhan Kim, Dongyoon Han, Sungryull Sohn, and Junmo Kim, "Facial Age Estimation via Extended Curvature Gabor Filter", IEEE International Conference on Image Processing (ICIP), 2015. [PDF]
Dongyoon Han and Junmo Kim "Unsupervised Orthogonal Basis Feature Selection", IEEE International Conference on Computer Vision and Pattern Recognition (CVPR), 2015. [PDF][CODE]
Dongyoon Han, Jung Eun, Pyunghwan Ahn, and Jeonghyo Ha, Donghoon Shin, Junmo Kim, “Automatic Drawing Simplification via Complex Zernike Moments”, ITC-CSCC, 2015.
Jiwhan Kim, Dongyoon Han, Yu-Wing Tai, and Junmo Kim, "Salient Region Detection via High Dimensional Color Transform", IEEE International Conference on Computer Vision and Pattern Recognition (CVPR), 2014. [PDF][Project Page]
Dongyoon Han, Jiwhan Kim, Jeongwoo Ju, Injae Lee, Jihun Cha, and Junmo Kim, "Efficient and Fast Multi-View Face Detection Based on Feature Transformation", ICACT, 2014.
Dongyoon Han and Junmo Kim, "Texture Classification Based on Discriminative Component Selection of Local Binary Pattern and Variants", FCV, 2014.
Wonjun Hwang, Jiwhan Kim, Dongyoon Han, and Junmo Kim, "A Survey of Face Recognition Technique for Real Media Processing", Korea Society Broadcast Engineers Magazine, Vol. 19, No. 3, pp. 111-122, Jul. 2014.