Donggyu Joo (주 동 규) [CV] [Scholar]

Research Scientists in Samsung Research.
​
Ph. D. in Electric Engineering
(Statistical Inference & Information Theory Lab at KAIST)​

Contact: jdg105@kaist.ac.kr 
‌

RESEARCH SKILLS

Machine Learning and Computer Vision with Deep Learning
- General Computer Vision Application
- Network Pruning & Network Compression
- Generative Adversarial Networks
‌- Network Architecture Design
‌- Object Classification

EXPERIENCES

Research Scientist, Samsung Research
- Sep. 2022 - Present
- In Visual Percetion Team, Gloal AI Center

EDUCATION

 - Ph.D,  School of Electric Engineering, KAIST, Korea
    Mar. 2018 - Aug. 2022
   Ph.D. Thesis: "Effective Deep Neural Network Compression Based on the Network Pruning"
   Advisor: Junmo Kim
   GPA: 4.23/4.30

- M.S,  School of Electric Engineering, KAIST, Korea
   Sep. 2016 – Feb. 2018
   M.S. Thesis: ''Gradual Net: Unconstrained Control of Feature Map SizeUsing Non-integer Strided Sampling''
   Advisor: Junmo Kim
   GPA: 4.21/4.30

- B.S,  School of Electric Engineering, KAIST, Korea
   Feb. 2013 – Aug. 2016
   GPA: 4.14/4.30. Summa-Cum-Laude

- K‌orea Science Academy of KAIST (KSA)
‌

PUBLICATION

Training Cartoonization Network without Cartoon
Doyeon Kim, Dongyeun Lee, Donggyu Joo, Junmo Kim
The 29th IEEE International Conference on Image Processing (ICIP '23)

Which Metrics For Network Pruning: Final Accuracy? or Accuracy Drop?
Donggyu Joo, Sunghun Baek, and Junmo Kim
The 29th IEEE International Conference on Image Processing (ICIP '22)

Global-Local Path Networks for Monocular Depth Estimation with Vertical CutDepth
Doyeon Kim, Woonghyun Ga, Pyungwhan Ahn, Donggyu Joo, Sehwan Chun, Junmo Kim
arXiv preprint arXiv:2201.07436

​Linear Combination Approximation of Feature for Channel Pruning
Donggyu Joo, Doyeon Kim, Eojindl Yi, and Junmo Kim
Efficient Deep Learning for Computer Vision CVPR Workshop 2022 (CVPRW '22)

Linearly Replaceable Filters for Deep Network Channel Pruning [PDF] [Code]
Donggyu Joo, Eojindl Yi, Sunghyun Baek, and Junmo Kim
35th AAAI Conference on Artificial Intelligence (AAAI '21) ​

Slimming Resnet by Slimming Shortcut [PDF]
Donggyu Joo, Doyeon Kim, and Junmo Kim
25th International Conference on Pattern Recognition (ICPR '20)

Delivering Meaningful Representation for Monocular Depth Estimation [PDF]
Doyeon Kim, Donggyu Joo, and Junmo Kim
25th International Conference on Pattern Recognition (ICPR '20 )

TiVGAN: Text to Image to Video Generation with Step-by-Step Evolutionary Generator [PDF]
Doyeon Kim*, Donggyu Joo*, and Junmo Kim (* equally contributed)
IEEE Access, Aug 2020.

Continual Learning with Extended Kronecker-factored Approximate Curvature [PDF]
Janghyun Lee, Hyeonggwon Hong, Donggyu Joo, and Junmo Kim.
IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR '20)

Residual Continual Learning [PDF]
Janghyun Lee, Donggyu Joo, Hyeonggwon Hong, and Junmo Kim
Thirty-Fourth AAAI Conference on Artificial Intelligence (AAAI'20) Oral Presentation

Unconstrained Control of Feature Map Size Using Non-integer Struded Sampling [PDF]
Donggyu Joo, Junho Yim, and Junmo Kim
2018 British Machine Vision Conference (BMVC '18)

Generating a Fusion Image: One' s Identity and Another’s Shape [PDF]
Donggyu Joo*, Doyeon Kim*, and Junmo Kim (* equally contributed)
IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR '18)

A gift from knowledge distillation: Fast optimization, network minimization and transfer learning [PDF]
Junho. Yim, Donggyu Joo, and Junmo Kim
IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR '17)

Identity Preserving Residual-Spatial Transformer Network
Donggyu Joo, Youngdong Kim, and Junmo Kim
The 5th International Conference on Robot Intelligence Technology and Applications (RiTA '17) Oral Presentation
​

HONORS AND AWARDS

‌- Young-Han Kim Global Leader Scholarship (2021)

- Winner of AI Reporter, AI WorldCup 2019. 
‌
- Winner of AI Reporter, AI WorldCup 2018.
‌
- Silver Prize (Signal Processing), 24th Samsung Humantech Paper Awards. (2018)
‌
- Silver Prize (Computer Science), 23th Samsung Humantech Paper Awards (2017)
‌
‌- Summa Cum Laude, School of Electrical Engineering, KAIST (2016)
‌
‌- Dean's List, School of Electrical Engineering, KAIST (Spring 2015)

PATENT

International

Method and Device for Learning Neural Network for Recognizing Class
Junmo Kim, Byungju Kim, Juchang Kim, Yeakang Lee, Minsuk Park, Juseung Yun, Jaeyoung Lee, Donggyu Joo
2018-09-04 / PCT/KR2018/010271

Domestic (KOR)

Deep Neural Network Compression and Pruning Method using Linearly Replaceable Filter
Junmo Kim, Donggyu Joo, Eojindl Yi, Sunghyun Baek, Daewon Kim, Doyeob Yeo, Cheolsig Pyo
2022-02-11 / 10-2022-0018275

Method and Device for Learning Neural Network for Recognizing Class
Junmo Kim, Byungju Kim, Juchang Kim, Yeakang Lee, Minsuk Park, Juseung Yun, Jaeyoung Lee, Donggyu Joo
2017-09-08 / 10-2017-0115451
‌
‌Knowledge Sharing Based Knowledge Transfer Method for Improving Quality of Knowledge and Apparatus Therefor
‌Junmo Kim, Junho Yim, Donggyu Joo, Jihoon Bae, Naesoo Kim, Chulsik Pyo
‌2017-09-13 / 10-2017-0117000
2019-12-31 / 10-2063376-0000
‌
‌Incremental Training Based Knowledge Transfer Method for Training Large Deep Neural Networks and Apparatus Therefor
‌‌Junmo Kim, Junho Yim, Donggyu Joo, Jihoon Bae, Naesoo Kim, Chulsik Pyo
‌2017-09-13 / 10-2017-0117001
‌2019-12-31 / 10-2063377-0000