A5015878311- Multimodal Machine Learning Applications
- Video Analysis and Summarization
- Human Pose and Action Recognition
- Domain Adaptation and Few-Shot Learning
- Advanced Image and Video Retrieval Techniques
- Anomaly Detection Techniques and Applications
- Visual Attention and Saliency Detection
- Machine Learning and ELM
- Agriculture, Soil, Plant Science
- Nutrition, Health and Food Behavior
- Korean Urban and Social Studies
- Agricultural Systems and Practices
- Advanced Neural Network Applications
- Video Surveillance and Tracking Methods
- Bayesian Methods and Mixture Models
- Marine and Coastal Research
- Global trade and economics
- Semiconductor materials and devices
- Handwritten Text Recognition Techniques
- Semiconductor materials and interfaces
- Garlic and Onion Studies
- Face and Expression Recognition
- International Development and Aid
- Vehicle License Plate Recognition
- Regional Development and Environment
Electronics and Telecommunications Research Institute
2022
Korea Advanced Institute of Science and Technology
2017-2021
Chonnam National University Hospital
2016
Weakly supervised object localization has recently attracted attention since it aims to identify both class labels and locations of objects by using image-level labels. Most previous methods utilize the activation map corresponding highest source. Exploiting only one probability is often biased into limited regions or sometimes even highlights background regions. To resolve these limitations, we propose use maps, named combinational maps (CCAM), which are linear combinations from lowest...
In this paper, we present a novel grid encoding model for content-aware image retargeting. contrast to previous approaches such as vertex-based and axis-aligned models, our approach takes each horizontal/vertical distance between two adjacent vertices an optimization variable. Upon difference-based scheme, every vertex position of target is subsequently determined after optimizing the one-dimensional values. Our quad edge-based has major advantages First, enables problem be developed in...
Domain adaptation (DA) is a representation learning methodology that transfers knowledge from label-sufficient source domain to label-scarce target domain. While most of early methods are focused on unsupervised DA (UDA), several studies semi-supervised (SSDA) recently suggested. In SSDA, small number labeled images given for training, and the effectiveness those data demonstrated by previous studies. However, SSDA approaches solely adopt embedding ordinary supervised losses, overlooking...
Haeju Lee, Oh Joon Kwon, Yunseon Choi, Minho Park, Ran Han, Yoonhyung Kim, Jinhyeon Youngjune Haebin Shin, Kangwook Kee-Eung Kim. Findings of the Association for Computational Linguistics: NAACL 2022.
Online action detection, which aims to identify an ongoing from a streaming video, is important subject in real-world applications. For this task, previous methods use recurrent neural networks for modeling temporal relations input sequence. However, these overlook the fact that image sequence includes not only of interest but background and irrelevant actions. This would induce units accumulate unnecessary information encoding features on interest. To overcome problem, we propose novel...
Recently, sports and ICT technology have been combined, enabling quantitative objective analysis of athlete competence. In the case soccer, competition athletes is underway in various companies, but due to technical limitations, many data are still being generated based on manual work experts. this paper, we propose an object motion recognition technique which a basis for further automation soccer analysis. We first classify objects game define recognizable each category. After that, design...
Recently, sports and ICT technology have been combined, enabling quantitative objective analysis of athlete competence. In the case soccer, competition athletes is underway in various companies, but due to technical limitations, many data are still being generated based on manual work experts. this paper, we propose an object motion recognition technique which a basis for further automation soccer analysis. We first classify objects game define recognizable each category. After that, design...
Improving farming activity competence of farm households has recently been considered one the most important factors for increasing income. However, few studies examine relationship between income and directly due to lack an available dataset. In this study, we household technical managerial based on nearly 30,000 consulting data gathered by Rural Development Administration, RDA. The major findings study are as follows: firstly, statistically significant differences in agricultural exist...
Due to the development of video understanding and big data analysis research field using deep learning technique, intelligent machines have replaced tasks that people performed in past various fields such as traffic, surveillance, security area.In sports field, especially soccer games, it is also attempting quantitative players games through or technique.However, because nature analysis, still difficult make sophisticated automatic due technical limitations.In this paper, we propose a based...
Weakly supervised object localization has recently attracted attention since it aims to identify both class labels and locations of objects by using image-level labels. Most previous methods utilize the activation map corresponding highest source. Exploiting only one probability is often biased into limited regions or sometimes even highlights background regions. To resolve these limitations, we propose use maps, named combinational maps (CCAM), which are linear combinations from lowest...
Most recent domain adaptation (DA) methods deal with unsupervised setup, which requires numerous target images for training. However, constructing a large-scale image set of the is occasionally much harder than preparing smaller number and label pairs. To cope problem, great attention recently paid to supervised (SDA), takes an extremely small amount labeled training (e.g., at most three examples per category). In SDA adapting deep networks towards very challenging due lack data, we tackle...
In preparation for Korea's participation in the CPTPP, this paper identifies issues of digital trade norms that have recently emerged, focusing on agricultural sector, by scrutinizing chapter and draws a response strategy. Although domestic smart agriculture is growing rapidly, size smart-farm equipment firms relatively small, government has focused only spread farms, so collection, management, protection, utilization farm data are not systematically or integratedly conducted. Digital...
Detecting text in natural scene images is a challenging task. In this paper, we propose character-level end-to-end detection algorithm images. general, tasks are categorized into three parts: localization, segmentation, and recognition. The proposed method aims not only to localize but also recognize text. To do these successfully, the consists of four steps: character candidate patch extraction, classification using ensemble ResNets, non-character region elimination, grouping via...