Minsu Cho

POSTECH, Computer Science and Engineering, Faculty Member

Followers

Following

Co-authors

Public Views

I am an assistant professor at the department of Computer Science and Engineering at POSTECH, South Korea, working as a faculty member of POSTECH Computer Vision Lab. My research lies in the areas of computer vision and machine learning, especially in the problems of object discovery, weakly-supervised learning, semantic correspondence, and graph matching.

Before joining POSTECH in the fall of 2016, I was a researcher (starting research position) in the Inria WILLOW team at École Normale Supérieure (ENS), Paris, France, where I have worked with Jean Ponce and closely collaborate with Cordelia Schmid in the Inria LEAR team. I completed my Ph.D. in 2012, under the supervision of Kyoung Mu Lee at Seoul National University, Korea. Web page: http://cvlab.postech.ac.kr/~mcho
Supervisors: Kyoung Mu Lee and Jean Ponce
Address: Dept. of Computer Science and Engineering
POSTECH (Pohang Univ. of Science and Technology, 포항공과대학교)
77 Cheongam-ro, Nam-gu, Pohang, Gyeongbuk, South Korea 37673

less

Michael Spivey

University of California, Merced

Giuseppe Primiero

Università degli Studi di Milano - State University of Milan (Italy)

Armando Marques-Guedes

UNL - New University of Lisbon

shuai ma

University of Edinburgh

John Sutton

Macquarie University

Eros Carvalho

Universidade Federal do Rio Grande do Sul

Gordana Dodig-Crnkovic

Chalmers University of Technology

Edwin Hancock

University of York

Zoran Zdravev

Univerzitet Goce Delcev

Miguel Gallegos

Universidad Nacional de Rosario

InterestsView All (7)

Uploads

Papers by Minsu Cho

Rebooting ACGAN: Auxiliary Classifier GANs with Stable Training

arXiv (Cornell University), Nov 1, 2021

Conditional Generative Adversarial Networks (cGAN) generate realistic images by incorporating cla... more Conditional Generative Adversarial Networks (cGAN) generate realistic images by incorporating class information into GAN. While one of the most popular cGANs is an auxiliary classifier GAN with softmax cross-entropy loss (ACGAN), it is widely known that training ACGAN is challenging as the number of classes in the dataset increases. ACGAN also tends to generate easily classifiable samples with a lack of diversity. In this paper, we introduce two cures for ACGAN. First, we identify that gradient exploding in the classifier can cause an undesirable collapse in early training, and projecting input vectors onto a unit hypersphere can resolve the problem. Second, we propose the Data-to-Data Cross-Entropy loss (D2D-CE) to exploit relational information in the class-labeled dataset. On this foundation, we propose the Rebooted Auxiliary Classifier Generative Adversarial Network (ReACGAN). The experimental results show that ReACGAN achieves state-of-the-art generation results on CIFAR10, Tiny-ImageNet, CUB200, and ImageNet datasets. We also verify that ReACGAN benefits from differentiable augmentations and that D2D-CE harmonizes with StyleGAN2 architecture. Model weights and a software package that provides implementations of representative cGANs and all experiments in our paper are available at .

Minsu Cho

Related Authors

Uploads

Papers by Minsu Cho

Log In