A5102964658- Advanced Image Processing Techniques
- Advanced Vision and Imaging
- Image Processing Techniques and Applications
- Machine Learning in Healthcare
- Gut microbiota and health
- Neural Networks and Applications
- Health, Environment, Cognitive Aging
- Topic Modeling
- Artificial Intelligence in Healthcare
- Image and Signal Denoising Methods
- Metabolomics and Mass Spectrometry Studies
- Image Enhancement Techniques
- User Authentication and Security Systems
- Natural Language Processing Techniques
- Dementia and Cognitive Impairment Research
- Robotics and Sensor-Based Localization
- Brain Tumor Detection and Classification
- Time Series Analysis and Forecasting
- Belt Conveyor Systems Engineering
- Multimodal Machine Learning Applications
- Advanced Algorithms and Applications
- Privacy, Security, and Data Protection
- Clostridium difficile and Clostridium perfringens research
- Semiconductor Lasers and Optical Devices
- Advanced Optical Sensing Technologies
SUNY Korea
2021-2024
Ajou University
2017-2023
State University of New York
2021
Samsung (South Korea)
2003-2015
This paper reviews the 2nd NTIRE challenge on single image super-resolution (restoration of rich details in a low resolution image) with focus proposed solutions and results. The had 4 tracks. Track 1 employed standard bicubic downscaling setup, while Tracks 2, 3 realistic unknown downgrading operators simulating camera acquisition pipeline. were learnable through provided pairs high train images. tracks 145, 114, 101, 113 registered participants, resp., 31 teams competed final testing...
Abstract Alzheimer’s disease (AD) is a progressive neurodegenerative condition marked by decline in cognitive functions with no validated modifying treatment. It critical for timely treatment to detect AD its earlier stage before clinical manifestation. Mild impairment (MCI) an intermediate between cognitively normal older adults and AD. To predict conversion from MCI probable AD, we applied deep learning approach, multimodal recurrent neural network. We developed integrative framework that...
The problem of enhancing the resolution a single low-resolution image has been popularly addressed by recent deep learning techniques. However, many approaches still fail to deal with extreme super-resolution scenarios because instability training. In this paper, we address issue adapting progressive scheme convolutional neural network. detail, overall training proceeds in multiple stages so that model gradually increases output resolution. our experiments, show property yields large...
Brain age estimation from anatomical features has been attracting more attention in recent years. This interest brain is motivated by the importance of biological prediction health informatics, with an application to early neurocognitive disorders. It well-known that normal aging follows a specific pattern, which enables researchers and practitioners predict human's its degeneration. In this paper, we model predicted cortical thickness data gathered large cohort images. We collected 2,911...
As large amounts of heterogeneous biomedical data become available, numerous methods for integrating such datasets have been developed to extract complementary knowledge from multiple domains sources. Recently, a deep learning approach has proposed with promising results in variety research areas. However, applying approaches dataset involves the strenuous design architectures. Thus, this paper, we present learning-based python package that can efficiently perform integration. The propose is...
Recent progress in deep learning-based models has improved photo-realistic (or perceptual) single-image super-resolution significantly. However, despite their powerful performance, many methods are difficult to apply real-world applications because of the heavy computational requirements. To facilitate use a model under such demands, we focus on keeping network efficient while maintaining its performance. In detail, design an architecture that implements cascading mechanism residual boost...
Image distortion classification and detection is an important task in many applications. For example when compressing images, if we know the exact location of distortion, then it possible to re-compress images by adjusting local compression level dynamically. In this paper, address problem detecting region classifying type a given image. We show that our model significantly outperforms state-of-the-art classifier, report accurate results for first time. expect such prove usefulness approach...
Natural language-based vehicle retrieval system makes controlling a city-scale traffic easy to maintain and adaptable changing requirements. It provides convenient means in managing flows or detecting accidents related specific vehicle. Such is different from most query-based video systems because the language for situations visible objects streams are limited. Existing techniques general problems measure similarity between representations representations. Our focuses on several features...
Investigators have studied the treatment effects on human health or disease, microbiome, and roles of microbiome disease. Especially, in a clinical trial, investigators commonly trace disease status over lengthy period to survey sequential progression for different groups (e.g., versus placebo, new old treatment). Hence, responses are often available form survival (i.e., time-to-event) stratified by groups. While recent web cloud platforms enabled user-friendly data processing analytics,...
We propose a novel method for identifying and classifying motions that offers significantly reduced computational cost as compared to deep convolutional neural network systems with comparable performance. Our new approach is inspired by the information processing architecture of biological visual systems, whereby spatial pyramid kernel features are efficiently extracted in real-time from temporally-differentiated image data. In this paper, we describe evaluate its performance hand motion...
Accessing medical data from multiple institutions is difficult owing to the interinstitutional diversity of vocabularies. Standardization schemes, such as common model, have been proposed solutions this problem, but schemes require expensive human supervision. This study aims construct a trainable system that can automate process semantic code mapping.To mapping between source and target codes, we compute embedding-based similarity corresponding descriptive sentences. We also implement...
The recent advance in massively parallel sequencing has enabled accurate microbiome profiling at a dramatically lowered cost. Then, the human been subject of intensive investigation public health and medicine. In meanwhile, researchers have developed lots data analysis methods, protocols, and/or tools. Among those, especially, web platforms can be highlighted because user-friendly interfaces streamlined protocols for long sequence analytic procedures. However, existing handle only...
Abstract Pairing (or blocking) is a design technique that widely used in comparative microbiome studies to efficiently control for the effects of potential confounders (e.g., genetic, environmental, or behavioral factors). Some typical paired (block) designs human are repeated measures profile each subject’s twice more than twice) (1) pre and post treatments see treatment on microbiome, (2) different organs body gut, mouth, skin) disparity between across) sites. Researchers have developed...
It has become common practice now to use random initialization schemes, rather than the pre-trained embeddings, when training transformer based models from scratch. Indeed, we find that word embeddings GloVe, and some sub-word extracted language such as T5 mT5 fare much worse compared initialization. This is counter-intuitive given well-known representational transfer-learning advantages of pre-training. Interestingly, also BERT mBERT better initialization, showing representations. In this...
Unmanned ground vehicles (UGVs) have emerged as a promising solution for reconnaissance missions, overcoming labor cost, frequency, and subjectivity issues associated with manual procedures. However, dynamic environments such construction sites, the constantly changing conditions hinder manager from planning UGV’s paths. For an autonomous monitoring mission, path should be dealt by having map site’s most recent scene. In this study, we develop mobile app capable of local updates overlaying...
High average output power over +3 dBm was obtained in 10 Gb/s electro-absorption modulated laser for transmission 80 km by amplifying the signals through monolithically integrated semiconductor optical amplifier, first time to our knowledge.
We address the problem of finding compact class sets for Korean font images taken under natural and noisy circumstances. are prone to misclassification due similar, yet subtly different visual characteristics. The classification becomes even more confusing when subject various pixel-wise or affine translations, such as scaling shear mapping. argue that many divisions inherently flawed in sense fonts divided an overly-fine manner. To tackle this issue, we propose a system discovers sets,...