A5009028758- Handwritten Text Recognition Techniques
- Advanced Image and Video Retrieval Techniques
- Advanced Neural Network Applications
- Image Retrieval and Classification Techniques
- Domain Adaptation and Few-Shot Learning
- Music and Audio Processing
- Antioxidant Activity and Oxidative Stress
- Speech Recognition and Synthesis
- Speech and Audio Processing
- Advanced Image Processing Techniques
- Plant biochemistry and biosynthesis
- Neural Networks and Applications
- Air Traffic Management and Optimization
- Innovation in Digital Healthcare Systems
- Internet of Things and Social Network Interactions
- Information and Cyber Security
- Image Processing Techniques and Applications
- Technology and Data Analysis
- Advanced Control and Stabilization in Aerospace Systems
- Aerospace Engineering and Applications
- Plant Reproductive Biology
- Plant nutrient uptake and metabolism
- Caching and Content Delivery
- IoT and Edge/Fog Computing
- Cloud Computing and Remote Desktop Technologies
Handong Global University
2014-2024
National Security Research Institute
2005-2016
Columbia University
2014
Seoul National University
2002-2008
Pohang University of Science and Technology
2007
Electronics and Telecommunications Research Institute
2005
NCSOFT (South Korea)
2003
Korea Advanced Institute of Science and Technology
1998-2002
Sewanee: The University of the South
2002
Here, we synthesized highly stable DNA-embedded Au/Ag core–shell nanoparticles (NPs) by a straightforward silver-staining of DNA-modified Au (AuNPs); unlike conventional DNA-surface modified NPs that present particle stability issues, offer an extraordinary with nanoscale controllability silver shell thickness; these show excellent biorecognition properties and Ag shell-thickness-based optical properties, distinctively different from those mixture AuNPs AgNPs or Ag/Au alloy nanoparticles.
We present a novel high-fidelity real-time neural vocoder called VocGAN.A recently developed GAN-based vocoder, MelGAN, produces speech waveforms in real-time.However, it often waveform that is insufficient quality or inconsistent with acoustic characteristics of the input mel spectrogram.VocGAN nearly as fast but significantly improves and consistency output waveform.VocGAN applies multi-scale generator hierarchically-nested discriminator to learn multiple levels properties balanced way.It...
To address the issue of catastrophic forgetting in neural networks, we propose a novel, simple, and effective solution called neuron-level plasticity control (NPC). While learning new task, proposed method preserves existing knowledge from previous tasks by controlling network at neuron level. NPC estimates importance value each consolidates important neurons applying lower rates, rather than restricting individual connection weights to stay close values optimized for tasks. The experimental...
This paper proposes a statistical character structure modeling method. It represents each stroke by the distribution of feature points. The is represented joint component strokes. In proposed model, relationship effectively reflected dependency. can represent all kinds in systematic way. Based on representation, neighbor selection method also proposed. measures importance mutual information among With such measure, important relationships are selected nth order probability approximation...
Recent efforts in learned cardinality estimation (CE) have substantially improved accuracy and query plans inside optimizers. However, achieving decent efficiency, scalability, the support of a wide range queries at same time, has remained questionable. Rather than falling back to traditional approaches trade off one criterion with another, we present new approach that achieves all these. Our method, called ASM, harmonizes autoregressive models for per-table statistics estimation, sampling...
Proposes some strategies to improve the recognition performance of a feature matching method for handwritten Chinese character (HCCR). Favorable modifications are given all stages throughout recognition. In pre-processing, we devised modified nonlinear normalization algorithm and connectivity-preserving smoothing algorithm. For extraction, an efficient directional decomposition systematic approach design blurring mask presented. Finally, LVQ3 is applied optimize reference vectors...
In capsule networks, the routing algorithm connects capsules in consecutive layers, enabling upper-level to learn higher-level concepts by combining of lower-level capsules. Capsule networks are known have a few advantages over conventional neural including robustness 3D viewpoint changes and generalization capability. However, some studies reported negative experimental results. Nevertheless, reason for this phenomenon has not been analyzed yet. We empirically effect five different...
We propose a multi-singer emotional singing voice synthesizer, Muse-SVS, that expresses emotion at various intensity levels by controlling subtle changes in pitch, energy, and phoneme duration while accurately following the score. To control multiple style attributes avoiding loss of fidelity expressiveness due to interference between attributes, Muse-SVS represents all their relations together joint embedding unified space. can express not included training data through interpolation...
Deep1 Convolutional Neural Network (DCNN) is a break-through technology in image recognition. However, because of extreme computing resource requirements, DCNN need to be implemented by hardware accelerator. In this paper, we present an FPGA-based accelerator design techniques for handwritten Hangul character recognition engine. We achieved about 11.9ms time per with Xilinx FPGA Our optimization was performed HLS and SDAccel environment targeting Kintex XCKU115 from Xilinx. outperforms CPU...
As the needs of education programming language increase, importance learning environment has been emphasized. Based on social trends, various web services for which improves computational thinking a student. Among services, block-based languages are well known as an effective educational tool primary student though freshman in university. These tools proven to be useful since provide same students and teachers anytime anywhere. However, environments may insufficient practical languages, such...
We propose a novel transfer learning framework for pathological image analysis, the Response-based Cross-task Knowledge Distillation (RCKD), which improves performance of model by pretraining it on large unlabeled dataset guided high-performance teacher model. RCKD first pretrains student to predict nuclei segmentation results images, and then fine-tunes pretrained downstream tasks, such as organ cancer sub-type classification region segmentation, using relatively small target datasets....