A5064858861- Video Surveillance and Tracking Methods
- Water Systems and Optimization
- Human Pose and Action Recognition
- Advanced Neural Network Applications
- Diverse Approaches in Healthcare and Education Studies
- Advanced Image and Video Retrieval Techniques
- Gait Recognition and Analysis
- Anomaly Detection Techniques and Applications
- Fire Detection and Safety Systems
- Educational Systems and Policies
- Water Quality Monitoring Technologies
- Radiomics and Machine Learning in Medical Imaging
- Multimodal Machine Learning Applications
- Diverse Topics in Contemporary Research
- Water Treatment and Disinfection
- Vehicle License Plate Recognition
- Video Analysis and Summarization
- Innovation in Digital Healthcare Systems
- Technology and Data Analysis
- AI in cancer detection
- IoT and GPS-based Vehicle Safety Systems
- Industrial Vision Systems and Defect Detection
- Data Management and Algorithms
- Surgical Simulation and Training
- Human Mobility and Location-Based Analysis
Electronics and Telecommunications Research Institute
2018-2024
Pohang University of Science and Technology
2013
Deep learning-based object detectors have driven no-table progress in multi-object tracking algorithms. Yet, current methods mainly focus on simple, regular motion patterns pedestrians or vehicles. This leaves a gap algorithms for targets with nonlinear, irregular motion, like athletes. Additionally, relying the Kalman filter recent falls short when defies its linear assumption. To over-come these issues, we propose novel online and robust approach named deep ExpansionIoU (Deep-EIoU), which...
The recent research by deep learning has shown many breakthroughs with high performance that were not achieved traditional machine algorithms. Particularly in the field of object detection, commercial products accuracy real environment are applied through methods. However, detection method using convolutional neural network (CNN) a disadvantage large number feature maps should be generated order to robust against scale change and occlusion object. Also, simply raising does improve...
In this paper, we proposes a novel drowsiness detection algorithm using camera near the dashboard. The proposed detects driver's face in image and estimates landmarks region. order to detect face, uses an AdaBoost classifier based on Modified Census Transform features. And regressing Local Binary Features for landmark detection. Eye states (closed, open) is determined by value of Aspect Ratio which easily calculated eye provides realtime performance that can be run embedded device. We...
In recent years, vision-based object detection methods using convolutional neural network (CNN) have been very successful. However, the method CNN feature has a disadvantage that lots of maps should be generated in order to robust against scale change and occlusion object. Also, simply raising large number does not improve performance. We propose multi-scale vehicle with spatial pyramid pooling which is by improving conventional YOLOv3 algorithm. The proposed was evaluated through UA-DETRAC...
In this paper, we propose a novel framework for multi-target multi-camera tracking (MTMCT) of vehicles based on metadata-aided re-identification (MA-ReID) and the trajectory-based camera link model (TCLM). Given video sequence corresponding frame-by-frame vehicle detections, first address isolated tracklets issue from single (SCT) by proposed traffic-aware single-camera (TSCT). Then, after automatically constructing TCLM, solve MTMCT MA-ReID. The TCLM is generated topological configuration...
The 1 <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">st</sup> Workshop on Maritime Computer Vision (MaCVi) 2023 focused maritime computer vision for Unmanned Aerial Vehicles (UAV) and Surface Vehicle (USV), organized several subchallenges in this domain: (i) UAV-based Object Detection, (ii) Mar-itime Tracking, (iii) USV-based Obstacle Segmentation (iv) Detection. were based the SeaDronesSee MODS benchmarks. This report summarizes main findings...
Multi-Object Tracking on humans has improved rapidly with the development of object detection and re-identification algorithms. However, multi-actor tracking over similar appearance non-linear movement can still be very challenging even for state-of-the-art algorithm. Current motion-based algorithms often use Kalman Filter to predict motion an object, however, its linear assumption cause failure in when target is not moving linearly. And multi-player sports field, because players same team...
Gait recognition, which refers to the recognition or identification of a person based on their body shape and walking styles, derived from video data captured distance, is widely used in crime prevention, forensic identification, social security. However, best our knowledge, most existing methods use appearance, posture temporal feautures without considering learned attention mechanism for global local information fusion. In this paper, we propose novel gait framework, called Temporal...
Multi-camera multiple people tracking has become an increasingly important area of research due to the growing demand for accurate and efficient indoor systems, particularly in settings such as retail, healthcare centers, transit hubs. We proposed a novel multi-camera method that uses anchor-guided clustering cross-camera re-identification spatio-temporal consistency geometry-based ID reassigning. Our approach aims improve accuracy by identifying key features are unique every individual...
Gait recognition is one of technology for biometrics at a distance that can be used to identify human through walking postures and body shape. In the field information forensics security, gait exploited crime prevention, forensic identification, social security. However, existing methods usually consider appearance, posture temporal separately, thus, we focus on considering learned attention mechanism fuse these features in global local manner. this article, novel framework Temporal...
Vehicle color recognition is one of the important part in ITS (Intelligent Transportation System). This paper presents a new vehicle classification technique for CCTV systems via representative region extraction and Convolutional Neural Net (CNN). The Harris corner point detection method used to generate probability map region. From map, are randomly selected an input image CNN. Finally, we trained CNN model with it. In order evaluate performance proposed method, acquired total 5,941 images...
The Radar Object Detection 2021 (ROD2021) Challenge, held in the ACM International Conference on Multimedia Retrieval (ICMR) 2021, has been introduced to detect and classify objects purely using an FMCW radar for autonomous driving applications. As a robust sensor all-weather conditions, rich information hidden radio frequencies, which can potentially achieve object detection classification. This insight will provide new perception solution vehicle even adverse scenarios. ROD2021 Challenge...
The 1$^{\text{st}}$ Workshop on Maritime Computer Vision (MaCVi) 2023 focused maritime computer vision for Unmanned Aerial Vehicles (UAV) and Surface Vehicle (USV), organized several subchallenges in this domain: (i) UAV-based Object Detection, (ii) Tracking, (iii) USV-based Obstacle Segmentation (iv) Detection. were based the SeaDronesSee MODS benchmarks. This report summarizes main findings of individual introduces a new benchmark, called Detection v2, which extends previous benchmark by...
Recently, a new noninvasive ultrasonic technique called the coronary Doppler vibrometry (CDV) was developed to detect audio-frequency vibrations in vessel wall and surrounding tissues generated by turbulence flow associated with stenosed artery. Inspiring clinical data for diagnosing artery stenosis (CAS) were obtained from CDV high sensitivity specificity. However, there still exists significant limitation, one of which is long examination time. In estimating myocardial vibrations, we must...
This paper focused on low-cost real-time driver's drowsiness detection by fusing facial image information obtained through the IR camera (Infrared Camera) and EEG (Electroencephalogram) signal acquired sensor. The proposed method was tested target board (i.MX6Quad). i.MX6Quad is SoCs (System-on-Chip) that integrate many processing units into one die, like main CPU, a video unit graphics for instance. Instead of RGB camera, applied to driver condition monitoring technology extracting feature...
Person Re-IDentification (ReID) is a pivotal method for pedestrian tracking and retrieval. This research inherently challenged by large changes in intra-class or small inter-class. To address this challenge, many researchers have recently introduced transformer-based models, which shown excellent results. The primary objective of these models to generate robust features that effectively distinguish between classes enable generalization. However, existing methods still suffer from class...
Segmenting lesions using brain MRI (Magnetic Resonance Imaging) images is a method that can help doctors determine the NIHSS (National Institutes of Health Stroke Scale) score. Therefore, we proposed SSNet model to segment foundation model. Our was constructed nnUNetv2, with good performance in medical image segmentation, and SAM-Tracker, model's IoU (Intersection over Union) 0.463, Dice 0.611, when only nnUNetv2 used, 0.569, 0.702. Although currently low, evolving, so it expected will be...