A5018432453- Video Surveillance and Tracking Methods
- Domain Adaptation and Few-Shot Learning
- Radiomics and Machine Learning in Medical Imaging
- Multimodal Machine Learning Applications
- AI in cancer detection
- Advanced Image and Video Retrieval Techniques
- Generative Adversarial Networks and Image Synthesis
- Human Pose and Action Recognition
- Image Enhancement Techniques
- Digital Radiography and Breast Imaging
- Advanced Neural Network Applications
- Advanced Vision and Imaging
- Visual Attention and Saliency Detection
- Retinal Diseases and Treatments
- Retinal Imaging and Analysis
- Glaucoma and retinal disorders
- Video Analysis and Summarization
- Infrared Thermography in Medicine
- Infrared Target Detection Methodologies
- Neural Networks and Applications
- Image Retrieval and Classification Techniques
- Topic Modeling
- Natural Language Processing Techniques
- Ophthalmology and Visual Impairment Studies
- Robotics and Automated Systems
Lutron Electronics (United States)
2019-2021
Pohang University of Science and Technology
2014-2018
Solutions Inc. (Japan)
2017
Naver (South Korea)
2016
Korea Post
2014
We propose a novel visual tracking algorithm based on the representations from discriminatively trained Convolutional Neural Network (CNN). Our pretrains CNN using large set of videos with ground-truths to obtain generic target representation. network is composed shared layers and multiple branches domain-specific layers, where domains correspond individual training sequences each branch responsible for binary classification identify in domain. train domain iteratively layers. When new...
We propose Dual Attention Networks (DANs) which jointly leverage visual and textual attention mechanisms to capture fine-grained interplay between vision language. DANs attend specific regions in images words text through multiple steps gather essential information from both modalities. Based on this framework, we introduce two types of for multimodal reasoning matching, respectively. The model allows attentions steer each other during collaborative inference, is useful tasks such as Visual...
The Visual Object Tracking challenge 2015, VOT2015, aims at comparing short-term single-object visual trackers that do not apply pre-learned models of object appearance. Results 62 are presented. number tested makes VOT 2015 the largest benchmark on tracking to date. For each participating tracker, a short description is provided in appendix. Features VOT2015 go beyond its VOT2014 predecessor are: (i) new dataset twice as large with full annotation targets by rotated bounding boxes and...
BackgroundMammography is the current standard for breast cancer screening. This study aimed to develop an artificial intelligence (AI) algorithm diagnosis of in mammography, and explore whether it could benefit radiologists by improving accuracy diagnosis.MethodsIn this retrospective study, AI was developed validated with 170 230 mammography examinations collected from five institutions South Korea, USA, UK, including 36 468 positive confirmed biopsy, 59 544 benign biopsy (8827 mammograms)...
We present an online visual tracking algorithm by managing multiple target appearance models in a tree structure. The proposed employs Convolutional Neural Networks (CNNs) to represent appearances, where CNNs collaborate estimate states and determine the desirable paths for model updates tree. By maintaining diverse branches of structure, it is convenient deal with multi-modality appearances preserve reliability through smooth along paths. Since share all parameters convolutional layers,...
Following the advance of style transfer with Convolutional Neural Networks (CNNs), role styles in CNNs has drawn growing attention from a broader perspective. In this paper, we aim to fully leverage potential improve performance general vision tasks. We propose Style-based Recalibration Module (SRM), simple yet effective architectural unit, which adaptively recalibrates intermediate feature maps by exploiting their styles. SRM first extracts information each channel pooling, then estimates...
Convolutional Neural Networks (CNNs) often fail to maintain their performance when they confront new test domains, which is known as the problem of domain shift. Recent studies suggest that one main causes this CNNs’ strong inductive bias towards image styles (i.e. textures) are sensitive changes, rather than contents shapes). Inspired by this, we propose reduce intrinsic style CNNs close gap between domains. Our Style-Agnostic (SagNets) disentangle encodings from class categories prevent...
Real-world image recognition is often challenged by the variability of visual styles including object textures, lighting conditions, filter effects, etc. Although these variations have been deemed to be implicitly handled more training data and deeper networks, recent advances in style transfer suggest that it also possible explicitly manipulate information. Extending this idea general problems, we present Batch-Instance Normalization (BIN) normalize unnecessary from images. Considering...
We propose a novel visual tracking algorithm based on the representations from discriminatively trained Convolutional Neural Network (CNN). Our pretrains CNN using large set of videos with ground-truths to obtain generic target representation. network is composed shared layers and multiple branches domain-specific layers, where domains correspond individual training sequences each branch responsible for binary classification identify in domain. train respect domain iteratively layers. When...
To develop an efficient deep neural network model that incorporates context from neighboring image sections to detect breast cancer on digital tomosynthesis (DBT) images.
We propose Dual Attention Networks (DANs) which jointly leverage visual and textual attention mechanisms to capture fine-grained interplay between vision language. DANs attend specific regions in images words text through multiple steps gather essential information from both modalities. Based on this framework, we introduce two types of for multimodal reasoning matching, respectively. The model allows attentions steer each other during collaborative inference, is useful tasks such as Visual...
Convolutional Neural Networks (CNNs) often fail to maintain their performance when they confront new test domains, which is known as the problem of domain shift. Recent studies suggest that one main causes this CNNs' strong inductive bias towards image styles (i.e. textures) are sensitive changes, rather than contents shapes). Inspired by this, we propose reduce intrinsic style CNNs close gap between domains. Our Style-Agnostic (SagNets) disentangle encodings from class categories prevent...
10519 Background: There is increasing interest in early detection of breast cancer by utilizing MRI high-risk populations. However, it still challenging to define and enrich the population. In this study, we developed an artificial intelligence (AI)-powered Imaging Biomarker Mammography (IBM) discover unique mammographic patterns, beyond simple density evaluations, that are related cancer. Methods: A total 49,577 mammography exams were collected develop AI-powered IBM, which 6,218 cancers....
Following the advance of style transfer with Convolutional Neural Networks (CNNs), role styles in CNNs has drawn growing attention from a broader perspective. In this paper, we aim to fully leverage potential improve performance general vision tasks. We propose Style-based Recalibration Module (SRM), simple yet effective architectural unit, which adaptively recalibrates intermediate feature maps by exploiting their styles. SRM first extracts information each channel pooling, then estimates...