Jiyun Jang
I am an undergraduate student in Computer Science & Engineering, Korea University.
I have done a research internship at VAI Lab in Korea University with Professor Jinkyu Kim. During that time, I conducted research on Domain Adaptation in LiDAR-based 3D object detection. While working on perception research, I became interested in how an agent plans its movements after perception.
I join the RIRO Lab at the Korea Advanced Institute of Science and Technology (KAIST) as an intern with Professor Daehyung Park.
My research interests include Imitation Learning, Deep Reinforcement Learning, Human-Robot Interaction(HRI).
Email /
CV /
Github /
Linkdin
|
|
Research
The goal of my research is to enable mobile manipulation robots to perform tasks in the real world with human-like capabilities.
Another objective is to enhance Human-Robot Interaction(HRI), allowing robots to integrate seamlessly into people's daily lives.
Currently, I am focused on controlling the behavior of a robot agent through Sample Efficient Robotic Reinforcement Learning (SERL).
|
|
Finetuning Pre-trained Model with Limited Data
for LiDAR-based 3D Object Detection by Bridging Domain Gaps
Jiyun Jang,
Mincheol Chang,
Jinkyu Kim
IROS, 2024 (Accepted)  
paper /
code /
(will be released)
We propose a novel method, called Domain Adaptive Distill-Tuning (DADT), to adapt a pre-trained model
with limited target data (≈100 LiDAR frames), retaining its
representation power and preventing it from overfitting.
Our experiments with LiDAR-based driving
benchmarks, such as the Waymo Open dataset and KITTI,
confirm that our method effectively finetunes a pre-trained
model, achieving significant gains in accuracy.
|
|
도메인 갭 해결을 통한 라이다 객체인식
모델 백본 특성 유지
장지윤,
장민철,
김진규
IEIE(대한전자공학회), 2024 (Accepted)   (Oral Presentation)
paper
|
Extra-curricular activities
|
- 2024 Korea University Silicon Valley Program: June 2024- July 2024
- Korea University AI club (AIKU) member: January 2023- December 2023
- Student President of Korea University College of Informatics: November 2021- November 2022
Website template from Jon Barron.
|
|