A5031220093- Advanced Vision and Imaging
- Robotics and Sensor-Based Localization
- Video Surveillance and Tracking Methods
- Advanced Neural Network Applications
- Advanced Image and Video Retrieval Techniques
- Human Pose and Action Recognition
- Autonomous Vehicle Technology and Safety
- Image Processing Techniques and Applications
- Tactile and Sensory Interactions
- Interactive and Immersive Displays
- Anomaly Detection Techniques and Applications
- Advanced MRI Techniques and Applications
- Industrial Vision Systems and Defect Detection
- COVID-19 diagnosis using AI
- Indoor and Outdoor Localization Technologies
- Traffic and Road Safety
- Image Retrieval and Classification Techniques
- Lung Cancer Diagnosis and Treatment
- Fault Detection and Control Systems
- Music Technology and Sound Studies
- Remote Sensing and LiDAR Applications
- Neural Networks and Applications
- Face and Expression Recognition
- Handwritten Text Recognition Techniques
- Hand Gesture Recognition Systems
Gwangju Institute of Science and Technology
2022-2023
International Graduate School of English
2022-2023
Gist (Czechia)
2022
Korea Advanced Institute of Science and Technology
2014-2021
Alzheimer's disease (AD) is one of the most common causes neurodegenerative affecting over 50 million people worldwide. However, AD diagnosis occurs in moderate to late stage, which means that optimal time for treatment has already passed. Mild cognitive impairment (MCI) an intermediate state between cognitively normal and patients. Therefore, accurate prediction conversion process MCI may allow patients start preventive intervention slow progression disease. Nowadays, neuroimaging...
Intelligent agents gather information and perceive semantics within the environments before taking on given tasks. The store collected in form of environment models that compactly represent surrounding environments. agents, however, can only conduct limited tasks without an efficient effective model. Thus, such model takes a crucial role for autonomy systems intelligent agents. We claim following characteristics versatile model: accuracy, applicability, usability, scalability. Although...
Accurately predicting pedestrian trajectories requires a human-like socio-physical understanding of movement, nearby pedestrians, and obstacles. However, traditional methods struggle to generate multiple in the same situation based on are computationally intensive, making real-time application difficult. To overcome these limitations, we propose SPU-BERT, fast multi-trajectory prediction model that incorporates two non-recursive BERTs for multi-goal (MGP) trajectory-to-goal (TGP). First, MGP...
Anomaly detection identifies anomaly samples that deviate significantly from normal patterns. Usually, the number of is extremely small compared to samples. To handle such imbalanced sample distribution, one-class classification has been widely used in identifying by modeling features data using only data. Recently, recurrent autoencoder (RAE) shown outstanding performance sequential other conventional methods. However, RAE, which a long-term dependency problem, optimized fixed-length...
Intelligent agents need to understand the surrounding environment provide meaningful services or interact intelligently with humans. The should perceive geometric features as well semantic entities inherent in environment. Contemporary methods general one type of information regarding at a time, making it difficult conduct high-level tasks. Moreover, running two types and associating resultant requires lot computation complicates software architecture. To overcome these limitations, we...
Surface mount technology (SMT) is a process for producing printed-circuit boards. The solder paste printer (SPP), package mounter, and reflow oven are used SMT. board on which the deposited from SPP monitored by inspector (SPI). If malfunctions due to defects, produces defective products, then abnormal patterns detected SPI. In this article, we propose convolutional recurrent reconstructive network (CRRN), decomposes anomaly generated SPI data. CRRN learns only normal data detects pattern...
An artificial intelligence (AI) model using chest radiography (CXR) may provide good performance in making prognoses for COVID-19.We aimed to develop and validate a prediction CXR based on an AI clinical variables predict outcomes patients with COVID-19.This retrospective longitudinal study included hospitalized COVID-19 at multiple medical centers between February 2020 October 2020. Patients Boramae Medical Center were randomly classified into training, validation, internal testing sets (at...
In this article, we introduce a new benchmark dataset for the challenging writing in air (WiTA) task—an elaborate task bridging vision and natural language processing (NLP). WiTA implements an intuitive method with finger movement human–computer interaction (HCI). Our will facilitate development of data-driven systems, which, thus, far have displayed unsatisfactory performance—due to lack as well traditional statistical models they adopted. consists five subdatasets two languages (Korean...
The automated home referred to as Smart Home is expected offer fully customized services its residents, reducing the amount of labor, thus improving human beings' welfare. Service robots and Internet Things (IoT) play key roles in development Home. service provision with these two main components a environment requires: 1) learning reasoning algorithms 2) integration robot IoT systems. Conventional computational intelligence-based do not successfully manage dynamic changes data, simple...
Accurate estimation of 3D geometry and camera motion enables a wide range tasks in robotics autonomous vehicles. However, the lack semantics performance degradation due to dynamic objects hinder its application real-world scenarios. To overcome these limitations, we design novel neural semantic visual odometry (VO) architecture on top simultaneous VO, object detection instance segmentation (SimVODIS) network. Next, propose an attentive pose with multi-task learning formulation for handling...
In the field of surface mount technology (SMT), early detection defects in production machines is crucial to prevent yield reduction. order detect machine without attaching additional costly sensors, attempts have been made classify solder paste printers using defective pattern (DSPP) images automatically obtained through inspection (SPI). However, since DSPP are sparse, various sizes, and hardly collected, existing CNN-based classifiers tend fail generalize over-fitted train set. Besides,...
We present a challenging dataset, ChangeSim, aimed at online scene change detection (SCD) and more. The data is collected in photo-realistic simulation environments with the presence of environmental non-targeted variations, such as air turbidity light condition changes, well targeted object changes industrial indoor environments. By collecting simulations, multi-modal sensor precise ground truth labels are obtainable RGB image, depth semantic segmentation, camera poses, 3D reconstructions....
Objective: With the recent development of various MRI-conditional cardiac implantable electronic devices (CIEDs), accurate identification and characterization CIEDs have become critical when performing MRI in patients with CIEDs.We aimed to develop evaluate a deep learning-based algorithm (DLA) that performs detection parameters, including safety, on chest radiograph (CR) single step compare its performance other related algorithms were recently developed. Materials Methods:We developed DLA...
Text entry aims to provide an effective and efficient pathway for humans deliver their messages computers. With the advent of mobile computing, recent focus text-entry research has moved from physical keyboards soft keyboards. Current keyboards, however, increase typo rate due a lack tactile feedback degrade usability devices large portion on screens. To tackle these limitations, we propose fully imaginary keyboard (I-Keyboard) with deep neural decoder (DND). The invisibility I-Keyboard...
Accurate change detection enables a wide range of tasks in visual surveillance, anomaly and mobile robotics. However, contemporary approaches assume an ideal matching between the current stored scenes, whereas only coarse is possible real-world scenarios. Thus, fail to show reported performance settings. To overcome this limitation, we propose SimSaC. SimSaC concurrently conducts scene flow estimation able detect changes with imperfect matches. train without additional manual labeling,...
Due to the change in life style and diet, modern people suffer from obesity, diabetes, other types of diseases. Regular practice exercise can alleviate negative effects diseases even cure certain cases. In addition, regular improves quality life. These facts have drawn much attention nowadays recognize importance exercise. As a result, more hope start exercising but they lack knowledge how what Professional counseling costs relatively expensive thus it is difficult for ordinary access...
In the field of robotics engineering and autonomous driving vehicles, precise estimation positions through visual place recognition (VPR) is crucial not only for reducing localization errors caused by odometry but also preventing creation ambiguous maps in unfamiliar environments. Despite numerous research efforts aimed at improving VPR performance addressing challenges such as illumination variation, occlusions, dynamic objects, contemporary approaches have primarily focused on model-based...
Scene Change Detection (SCD) is vital for applications such as visual surveillance and mobile robotics. However, current SCD methods exhibit a bias to the temporal order of training datasets limited performance on unseen domains; coventional benchmarks are not able evaluate generalization or consistency. To tackle these limitations, we introduce Generalizable Framework (GeSCF) in this work. The proposed GeSCF leverages localized semantics foundation model without any re-training fine-tuning...