Jin Young Choi

Seoul National University

Author Statistics

Papers

Citation

H-Index

i-10 index

Research Trends

Author Order

Document Type

Co-Authors

Hyung Jin Chang

University of Birmingham

Sangdoo Yun

Naver (South Korea)

Jongwon Choi

Chung-Ang University

Kimin Yun

Electronics and Telecommunications Research Institute

Byeongho Heo

Naver (South Korea)

Hawook Jeong

Jeju TechnoPark

Kwang Moo Yi

University of British Columbia

Soo Wan Kim

Chonnam National University Hospital

Ho Seok Ahn

University of Auckland

Youngjoon Yoo

Chung Cheong University

Cooperative Institutions

Seoul National University

128

Chinese Academy of Sciences

Samsung (South Korea)

University of Chinese Academy of Sciences

Korea University

Korea Advanced Institute of Science and Technology

Dalian University of Technology

Sun Yat-sen University

Shanghai Jiao Tong University

Harbin Institute of Technology

Author Statistics

Papers

Citation

H-Index

i-10 index

Research Field

Heterogeneous Double-Head Ensemble for Deep Metric Learning

IEEE Access (2020)

Youngmin Ro Jin Young Choi

The structure of a multi-head ensemble has been employed by many algorithms in various applications including deep metric learning. However, their structures have been empirically designed in a simple way such as using the same head structure, which leads to a limited ensemble effect due to lack of head diversity. In this paper, for an elaborate design of the multi-head ensemble structure, we establish design concepts based on three structural factors: designing the feature layer for extracting the ensemble-favorable feature vector, designing the shared part for memory savings, and designing the diverse multi-heads for performance improvement. Through rigorous evaluation of variants on the basis of the design concepts, we propose a heterogeneous double-head ensemble structure that drastically increases ensemble gain along with memory savings. In verifying experiments on image retrieval datasets, the proposed ensemble structure outperforms the state-of-the-art algorithms by margins of over 5.3%, 6.1%, 5.9%, and 1.8% in CUB-200, Car-196, SOP, and Inshop, respectively.

Ensemble Learning

Feature (linguistics)

Ensemble forecasting

10.1109/access.2020.3004579

Cite

Citations (0)

Over-Fit: Noisy-Label Detection based on the Overfitted Model Property.

arXiv (Cornell University) (2021)

Seulki Park Dae Ung Jo Jin Young Choi

Due to the increasing need to handle the noisy label problem in a massive dataset, learning with noisy labels has received much attention in recent years. As a promising approach, there have been recent studies to select clean training data by finding small-loss instances before a deep neural network overfits the noisy-label data. However, it is challenging to prevent overfitting. In this paper, we propose a novel noisy-label detection algorithm by employing the property of overfitting on individual data points. To this end, we present two novel criteria that statistically measure how much each training sample abnormally affects the model and clean validation data. Using the criteria, our iterative algorithm removes noisy-label samples and retrains the model alternately until no further performance improvement is made. In experiments on multiple benchmark datasets, we demonstrate the validity of our algorithm and show that our algorithm outperforms the state-of-the-art methods when the exact noise rates are not given. Furthermore, we show that our method can not only be expanded to a real-world video dataset but also can be viewed as a regularization method to solve problems caused by overfitting.

Overfitting

Noisy data

Benchmark (surveying)

Regularization

Deep Neural Networks

Source

Cite

Citations (1)

A Comparison Study of Input ESD Protection Schemes Utilizing NMOS, Thyristor, and Diode Devices

Communications and Network (2010)

Jin Young Choi

For three fundamental input-protection schemes suitable for high-frequency CMOS ICs, which utilize protection devices such as NMOS transistors, thyristors, and diodes, we attempt an in-depth comparison on HBM ESD robustness in terms of lattice heating inside protection devices and peak voltages developed across gate oxides in input buffers, based on DC, mixed-mode transient, and AC analyses utilizing a 2-dimensional device simulator. For this purpose, we construct an equivalent circuit model of input HBM test environments for CMOS chips equipped with input ESD protection circuits, which allows mixed-mode transient simulations for various HBM test modes. By executing mixed-mode simulations including up to six active protection devices in a circuit, we attempt a detailed analysis on the problems, which can occur in real tests. In the procedure, we suggest to a recipe to ease the bipolar trigger in the protection devices and figure out that oxide failure in internal circuits is determined by the peak voltage developed in the later stage of discharge, which corresponds to the junction breakdown voltage of the NMOS structure residing in the protection devices. We explain strength and weakness of each protection scheme as an input ESD protection circuit for high-frequency ICs, and suggest valuable guidelines relating design of the protection devices and circuits.

Overvoltage

Snapback

Transient voltage suppressor

10.4236/cn.2010.21002

Cite

Citations (7)

Robot arm trajectory planning using missile guidance algorithm

Society of Instrument and Control Engineers of Japan (2003)

D. Chwa J.J. Kang Ki Hong Im Jin Young Choi

This paper presents a novel approach to online trajectory planning of robot arms for the interception of a fast-maneuvering object under torque and velocity constraints. Especially, the trajectory planning in the Y-direction employs a missile guidance algorithm to intercept the maneuvering object. The proposed method yields that a robot arm can smoothly grasp a maneuvering object. Simulation results are included to demonstrate the proposed method.

Source

Cite

Citations (4)

A new discriminant analysis for non-normally distributed data based on datawise formulation of scatter matrices

Myoung Soo Park Jin Young Choi

In this paper, we propose a new discrminant analysis based on datawise formulation of scatter matrices to deal with the data of non-normal distribution. Starting from original LDA, datawise formulation of scatter matrices is derived and its meaning is clarified. Based on this formulation, a new feature extraction algorithm is presented. In this formulation, assumption on distribution of data is no more necessary, so appropriate feature space can be found from the data whose distribution is non-normal, as well as multimodally normal. Limitation on the feature dimension also can be removed, and by replacing the inverse matrix of within-class scatter matrix with especially assigned weights, computational problems originating from matrix inversion of within-scatter matrix can be fundamentally avoided. As a result, good feature space for classification task can be found without the problems of LDA. Performance of this algorithm has been evaluated by using feature for real classification tasks.

Scatter matrix

Matrix (chemical analysis)

Feature vector

Feature (linguistics)

10.1109/ijcnn.2009.5178841

Cite

Citations (2)

Rotation-Aware 3D Vehicle Detection From Point Cloud

IEEE Access (2021)

Hyun-Jun Choi Jiyeoup Jeong Jin Young Choi

Three-dimensional vehicle detection using LiDAR point clouds is important for the stability of autonomous driving. It can provide high-quality three-dimensional information for most obstacles. Although efficient algorithms based on a Bird's Eye View (BEV) feature map have been developed, many research issues still remain; in particular, the existing methods show a limited accuracy in estimating the rotation angle of a 3D object. In this paper, to improve the accuracy of rotation angle estimation, we propose a rotation-aware 3D vehicle detector that extracts distinguishable features from the proposals with various angles of rotations. Experiments are conducted on KITTI dataset and Waymo Open dataset. Our approach improves detection accuracy as well as rotation angle estimation accuracy against the existing algorithms without much loss of computational efficiency.

Feature (linguistics)

10.1109/access.2021.3095525

Cite

Citations (1)

Situation-aware Framework for Moving Object Detection in Moving Cameras

Kimin Yun Jongin Lim Jin Young Choi

10.1049/cp.2017.0146

Cite

Citations (0)

Motion-aware ensemble of three-mode trackers for unmanned aerial vehicles

Machine Vision and Applications (2021)

Kyuewang Lee Hyung Jin Chang Jongwon Choi Byeongho Heo Aleš Leonardis

BitTorrent tracker

Tracking (education)

Benchmark (surveying)

Mode (computer interface)

Homography

10.1007/s00138-021-01181-x

Cite

Citations (2)

A Thyristor-Only Input ESD Protection Scheme for CMOS RF ICs

Circuits and Systems (2011)

Jin Young Choi Choongkoo Park

We propose an input protection scheme composed of thyristor devices only without using a clamp NMOS device in order to minimize the area consumed by a pad structure in CMOS RF ICs. For this purpose, we suggest low-voltage triggering thyristor protection device structures assuming usage of standard CMOS processes, and attempt an in-depth comparison study with a conventional thyristor protection scheme incorporating a clamp NMOS device. The comparison study mainly focuses on robustness against the HBM ESD in terms of peak voltages applied to gate oxides in an input buffer and lattice heating inside protection devices based on DC and mixed-mode transient analyses utilizing a 2-dimensional device simulator. We constructed an equivalent circuit for the input HBM test environment of the CMOS chip equipped with the input ESD protection devices. And by executing mixed-mode simulations including up to four protection devices and analyzing the results for five different test modes, we attempt a detailed analysis on the problems which can occur in real HBM tests. We figure out strength of the proposed thyristor-only protection scheme, and suggest guidelines relating the design of the protection devices and circuits.

MOS-controlled thyristor

Human-body model

10.4236/cs.2011.23025

Cite

Citations (2)

Unified optimization framework for localization and tracking of multiple targets with multiple cameras

Computer Vision and Image Understanding (2017)

Moonsub Byeon Haanju Yoo Kikyung Kim Songhwai Oh Jin Young Choi

Tracking (education)

10.1016/j.cviu.2017.10.009

Cite

Citations (7)