Welcome to Hyeonggeun's Terminal!
3/27/2024 2:59:58 PM


$>

Harooo Communication
#other #ongoing

I am one of the project leader of Harooo Communication. A representative service is SNS 'Harooo' where writings last 24 hours. Harooo Communication was announced by media, and members of Harooo Communication are supported by NAVER.


High-Reliability Teleoperation Assistance based on Deep Learning and Human-Computer Interaction
#research #ongoing
  • Younggeol Cho, Hyeonggeun Yun, Jinwon Lee, Arim Ha, and Jihyeok Yun

    Due to the limitations of autonomous driving technology, teleoperation is used extensively in hazardous environments such as military operations. However, the performance of teleoperated driving is primarily influenced by the driver's skill level. In other words, unskilled drivers need extensive training for teleoperation in harsh and unusual environments, such as off-road. In this letter, we propose GoonDAE, a novel denoising-based real-time driver assistance method that enables stable teleoperated off-road driving. We introduce a denoising autoencoder (DAE) based on a skip-connected long short-term memory (LSTM) to assist the unskilled driver control input through denoising. In this approach, it is assumed that the control input of an unskilled driver is equivalent to that of a skilled driver with noise. We train GoonDAE using the skilled driver control inputs and sensor data collected from our simulated off-road driving environment. Our experiments in the simulated off-road environment show that GoonDAE significantly improves the driving stability of unskilled drivers.

  • Hyeonggeun Yun, Younggeol Cho, Jinwon Lee, Arim Ha, and Jihyeok Yun

    Unmanned ground vehicles (UGVs) in unstructured environments mostly operate through teleoperation. To enable stable teleoperated driving in unstructured environments, some research has suggested driver assistance and evaluation methods that involve user studies, which can be costly and require lots of time and effort. A simulation model-based approach has been proposed to complement the user study; however, the models on teleoperated driving do not account for unstructured environments. Our proposed solution involves simulation models of teleoperated driving for drivers that utilize a deep generative model. Initially, we build a teleoperated driving simulator to imitate unstructured environments based on previous research and collect driving data from drivers. Then, we design and implement the simulation models based on a conditional variational autoencoder (CVAE). Our evaluation results demonstrate that the proposed teleoperated driving model can generate data by simulating the driver appropriately in unstructured canyon terrains.

  • Hyeonggeun Yun, Younggeol Cho, Arim Ha, and Jihyeok Yun

    This paper presents design guidelines for teleoperated driving interfaces within computational driver assistance systems for unstructured environments. The purpose of the guidelines is to manage the unpredictability of computational model-based assistance in unstructured environments in order to reduce user workload. Thus, we conducted a user study to evaluate workload and obtain insights into both the advantages and disadvantages of the computational driver assistance system in order to develop the guidelines. The study utilized a deep learning-based driver assistance method in simulated environments to observe the workload of users while teleoperated driving with the assistance method. Based on the user study, we proposed guidelines for teleoperated driving interface with computational driver assistance systems. We anticipate that the proposed guidelines could improve the understanding of computational driver assistance systems and reduce the workload of teleoperated driving in unstructured environments, thereby enhancing driver’s trust as well as comfort.

  • Hyeonggeun Yun

    Teleoperated driving is a leading approach for operating unmanned ground vehicles (UGVs) in unstructured environments, where driving stability is crucial. However, time delays may compromise this stability. In this letter, we propose a computational delay compensation method based on deep learning models to address the control transmission delay. Initially, we collect teleoperated driving data from simulated unstructured environments and then design a delay compensation method utilizing time-series forecasting models. This method generates future control inputs equivalent to the delayed time steps. Our evaluation demonstrates the possibility of our delay compensation method for teleoperated driving in unstructured environments.


NOTTO
#others #finished

I developed an iOS application named "NOTTO", that manages what not to do with emojis and widgets using Swift language and Firebase database.


Chatbot with Touch and Graphics: An Interaction of Users for Emotional Expression and Turn-taking
#research #finished
  • Hyeonggeun Yun, Auejin Ham, Jin Kim, Taeyeong Kim, Jeongeun Kim, Haechan Lee, Jongrae Park, Jinkyu Jang

    Use of chatbots for emotional exchange is recently increasing in various domains. However, as existing chatbots have been considered in terms of natural language processing techniques for interaction with text-based chatting, there are problems with the flow and the style of the conversation. Consequently, chatbot interaction with users is lacking in terms of considering the emotions of users and managing turn-taking in conversation. We suggest a new interaction technique having touch interactions with graphic interfaces (TwG) to solve these problems. In the proposed system, users send their emotions and manage turn-taking through TwG technique. This project is a part of DGIST UGRP (Undergraduate Group Research Program), and the research is supported by DGIST UGRP Grant.


Effects of Human-like Appearance and Non-Speech Sound on Psychological Resistance on SmartSpeaker
#research #finished
  • Gyeongbin Park, Hyeonggeun Yun, ChaeYeon Bang, Hyounjung Kang, Jinkyu Jang

    To help language learners achieve fluency, instructors often focus on teaching formulaic sequences (FS)–phrases such as idioms or phrasal verbs that are processed, stored and retrieved holistically. Teaching FS effectively is challenging as it heavily involves instructors’ intuition, prior knowledge, and manualefforts to identify a set of FSs with high utility. We suggest FSIST, a tool that supports instructorsfor video-based instruction of FS. The core idea of FSIST is to utilize videos at scale to build a list of FSs along with videos that include example usages. To evaluate how FSIST can effectively supportinstructors, we conducted a user study with three English instructors. Results show that the browsing interactions provided in FSIST support instructors to efficiently find parts of videos that show example usages of FSs.


Supporting Instruction of Formulaic Sequences Using Videos at Scale
#research #finished
  • Kyung Je Jo, Hyeonggeun Yun, Juho Kim

    To help language learners achieve fluency, instructors often focus on teaching formulaic sequences (FS)–phrases such as idioms or phrasal verbs that are processed, stored and retrieved holistically. Teaching FS effectively is challenging as it heavily involves instructors’ intuition, prior knowledge, and manualefforts to identify a set of FSs with high utility. We suggest FSIST, a tool that supports instructorsfor video-based instruction of FS. The core idea of FSIST is to utilize videos at scale to build a list of FSs along with videos that include example usages. To evaluate how FSIST can effectively supportinstructors, we conducted a user study with three English instructors. Results show that the browsing interactions provided in FSIST support instructors to efficiently find parts of videos that show example usages of FSs.


Mine Yourself!: A Role-playing Privacy Tutorial in Virtual Reality Environment
#research #finished
  • Junsu Lim*, Hyeonggeun Yun*, Auejin Ham*, and Sunjun Kim. *Authors are equally contributed

    Virtual Reality (VR) has potential vulnerabilities in privacy risks from collecting a wide range of data with higher density. Various designs to provide information on Privacy Policy (PP) have improved the awareness and motivation towards privacy risks. However, most of them have focused on desktop environments, not utilizing the full potential of VR’s immersive interactivity. Therefore, we proposed a role-playing mechanism to provide an immersive experience of interacting with PP’s key entities. First, our formative study found insights for PP where VR users had limited awareness of what data to be collected and how to control them. Following this, we implemented a VR privacy tutorial based on our role-playing mechanism and PP from off-the-shelf VR devices. Our privacy tutorial increased a similar amount of privacy awareness with significantly higher satisfaction (p=0.007) than conventional PP. Our tutorial also showed marginally higher usability (p=0.11).


Interaction-based explainable AI
#research #finished

I developed a web-based system that users can modify the images by interactions to understand image classification models using React library and FastAPI framework. I submitted an academic paper to Explainable AI for Computer Vision (XAI4CV) Workshop at CVPR 2024.


Self Design Tool for Web
#other #finished

I developed self design tool for web. Users can make and modify their poster, logo, and brand design using the tool. This is a work from Oasyss Story.


DGIST Tutoring
#other #finished

Since 2018, I have been an undergraduate tutor of DGIST. In 2018, I worked as a tutor of General Chemistry I in the first semester and Multivariable Calculus in the second semester. In 2019, I worked as a tutor of Linear Algebra in the first semester and Object-Oriented Programmin in the second semester. Now, I am working as a tutor of Data Structure for this semester. While I have been working as a tutor, I try to fill the gaps which I did not understand when I was a fresher student. Thanks to the efforts, I was selected as a superb tutor in the first semester of 2018, and the second semester of 2019.