• IEMEK Journal of Embedded Systems and Applications
• /
• v.17 no.5
• /
• pp.273-280
• /
• 2022

Along with the advancement of deep learning technology, securing high-quality dataset for verification of developed technology is emerging as an important issue, and developing robust deep learning models to the domestic road environment is focused by many research groups. Especially, unlike expressways and automobile-only roads, in the complex city driving environment, various dynamic objects such as motorbikes, electric kickboards, large buses/truck, freight cars, pedestrians, and traffic lights are mixed in city road. In this paper, we built our dataset through multi camera-based processing (collection, refinement, and annotation) including the various objects in the city road and estimated quality and validity of our dataset by using YOLO-based model in object detection. Then, quantitative evaluation of our dataset is performed by comparing with the public dataset and qualitative evaluation of it is performed by comparing with experiment results using open platform. We generated our 2D dataset based on annotation rules of KITTI/COCO dataset, and compared the performance with the public dataset using the evaluation rules of KITTI/COCO dataset. As a result of comparison with public dataset, our dataset shows about 3 to 53% higher performance and thus the effectiveness of our dataset was validated.

• Database Research
• /
• v.34 no.3
• /
• pp.86-98
• /
• 2018

Point-of-Interest(POI) recommendation systems suggest the most interesting POIs to users considering the current location and time. With the rapid development of smartphones, internet-of-things, and location-based social networks, it has become feasible to accumulate huge amounts of user POI visits. Therefore, instant recommendation of interesting POIs at a given time is being widely recognized as important. To increase the performance of POI recommendation systems, several studies extracting users' POI sequential preference from POI check-in data, which is intended for implicit feedback, have been suggested. However, when constructing a model utilizing sequential preference, the model encounters possibility of data distortion because of a low number of observed check-ins which is attributed to intensified data sparsity. This paper suggests refinement of temporal intervals based on data confidence. When building a POI recommendation system using temporal intervals to model the POI sequential preference of users, our methodology reduces potential data distortion in the dataset and thus increases the performance of the recommendation system. We verify our model's effectiveness through the evaluation with the Foursquare and Gowalla dataset.

• ETRI Journal
• /
• v.39 no.1
• /
• pp.97-107
• /
• 2017

A fast and accurate building-level visual place recognition method built on an image-retrieval scheme using street-view images is proposed. Reference images generated from street-view images usually depict multiple buildings and confusing regions, such as roads, sky, and vehicles, which degrades retrieval accuracy and causes matching ambiguity. The proposed practical database refinement method uses informative reference image and keypoint selection. For database refinement, the method uses a spatial layout of the buildings in the reference image, specifically a building-identification mask image, which is obtained from a prebuilt three-dimensional model of the site. A global-positioning-system-aware retrieval structure is incorporated in it. To evaluate the method, we constructed a dataset over an area of $0.26km^2$. It was comprised of 38,700 reference images and corresponding building-identification mask images. The proposed method removed 25% of the database images using informative reference image selection. It achieved 85.6% recall of the top five candidates in 1.25 s of full processing. The method thus achieved high accuracy at a low computational complexity.

• Journal of Information Processing Systems
• /
• v.17 no.2
• /
• pp.385-398
• /
• 2021

The complexity of deep learning models affects the real-time performance of gesture recognition, thereby limiting the application of gesture recognition algorithms in actual scenarios. Hence, a residual learning neural network based on a deep convolutional neural network is proposed. First, small convolution kernels are used to extract the local details of gesture images. Subsequently, a shallow residual structure is built to share weights, thereby avoiding gradient disappearance or gradient explosion as the network layer deepens; consequently, the difficulty of model optimisation is simplified. Additional convolutional neural networks are used to accelerate the refinement of deep abstract features based on the spatial importance of the gesture feature distribution. Finally, a fully connected cascade softmax classifier is used to complete the gesture recognition. Compared with the dense connection multiplexing feature information network, the proposed algorithm is optimised in feature multiplexing to avoid performance fluctuations caused by feature redundancy. Experimental results from the ISOGD gesture dataset and Gesture dataset prove that the proposed algorithm affords a fast convergence speed and high accuracy.

• Journal of the Korea Society of Computer and Information
• /
• v.27 no.12
• /
• pp.77-84
• /
• 2022

In this paper, we propose a Label Reassignment Strategy to improve the performance of an object detection algorithm. Our approach involves two stages: an inference stage and an assignment stage. In the inference stage, we perform multi-scale inference with predefined scale sizes on a trained model and re-infer masked images to obtain robust classification results. In the assignment stage, we calculate the IoU between bounding boxes to remove duplicates. We also check box and class occurrence between the detection result and annotation label to re-assign the dominant class type. We trained the YOLOX-L model with the re-annotated dataset to validate our strategy. The model achieved a 3.9% improvement in mAP and 3x better performance on AP_S compared to the model trained with the original dataset. Our results demonstrate that the proposed Label Reassignment Strategy can effectively improve the performance of an object detection model.

• Journal of Korea Multimedia Society
• /
• v.15 no.5
• /
• pp.624-631
• /
• 2012

For many medical imaging systems, volume datasets are stored as a compressed form, so that the dataset has to be decompressed before it is visualized. Since the decompression process takes quite a long time, we present an acceleration method for medical volume decompression using GPU. Our method supports that both lossy and lossless compression and progressive refinement is possible to satisfy variable user requirements. Moreover, our decompression method is well parallelized for GPU so that the decompression takes a very short time. Finally, we designed that the decompression and volume rendering work in one framework so that the selective decompression is available. As a result, we gained additional improvement in volume decompression.

• Journal of the Korea Society of Computer and Information
• /
• v.28 no.4
• /
• pp.53-63
• /
• 2023

The utilization of department information in bibliometric analysis using scientific and technological literature is highly advantageous. In this paper, the department information dataset was built through the screening, data refinement, and classification processing of authors' department type belonging to university institutions appearing in academic journals in the field of science and technology published in Korea, and the automatic classification model based on deep learning was developed using the department information dataset as learning data and verification data. In addition, we analyzed the co-authorship structure and network in the field of computer science using the department information dataset and affiliation information of authors from domestic academic journals. The research resulted in a 98.6% accuracy rate for the automatic classification model using Korean department information. Moreover, the co-authorship patterns of Korean researchers in the computer science and engineering field, along with the characteristics and centralities of the co-author network based on institution type, region, institution, and department type, were identified in detail and visually presented on a map.

• Proceedings of the KSRS Conference
• /
• 2003.11a
• /
• pp.525-527
• /
• 2003

Recently, the spatial resolution of earth observation satellites is significantly increased to a few meters. Such high spatial resolution images definitely will provide lots of information for detail-thirsty remote sensing users. However, it is more difficult to develop automated image algorithms for automated image feature extraction and pattern recognition. In this study, we propose a two-stage procedure to extract road information from high resolution satellite images. At first stage, a watershed segmentation technique is developed to classify the image into various regions. Then, a knowledge is built for road and used to extract the road regions. In this study, we use panchromatic and multi-spectral images of the IKONOS satellite as test dataset. The experiment result shows that the proposed technique can generate suitable and meaningful road objects from high spatial resolution satellite images. Apparently, misclassified regions such as parking lots are recognized as road needed further refinement in future research.

• Health Policy and Management
• /
• v.25 no.3
• /
• pp.197-206
• /
• 2015

Background: The purpose of this study was to propose a method for developing a measure of hospital-wide all-cause risk-standardized readmissions using administrative claims data in Korea and to discuss further considerations in the refinement and implementation of the readmission measure. Methods: By adapting the methodology of the United States Center for Medicare & Medicaid Services for creating a 30-day readmission measure, we developed a 6-step approach for generating a comparable measure using Korean datasets. Using the 2010 Korean National Health Insurance (NHI) claims data as the development dataset, hierarchical regression models were fitted to calculate a hospital-wide all-cause risk-standardized readmission measure. Six regression models were fitted to calculate the readmission rates of six clinical condition groups, respectively and a single, weighted, overall readmission rate was calculated from the readmission rates of these subgroups. Lastly, the case mix differences among hospitals were risk-adjusted using patient-level comorbidity variables. The model was validated using the 2009 NHI claims data as the validation dataset. Results: The unadjusted, hospital-wide all-cause readmission rate was 13.37%, and the adjusted risk-standardized rate was 10.90%, varying by hospital type. The highest risk-standardized readmission rate was in hospitals (11.43%), followed by general hospitals (9.40%) and tertiary hospitals (7.04%). Conclusion: The newly developed, hospital-wide all-cause readmission measure can be used in quality and performance evaluations of hospitals in Korea. Needed are further methodological refinements of the readmission measures and also strategies to implement the measure as a hospital performance indicator.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.19 no.1
• /
• pp.189-196
• /
• 2024

Frequent false positives alarm from the Intelligent Selective Control System have raised significant concerns. These persistent issues have led to declines in operational efficiency and market credibility among agents. Developing a new model or replacing the existing one to mitigate false positives alarm entails substantial opportunity costs; hence, improving the quality of the training dataset is pragmatic. However, smaller organizations face challenges with inadequate capabilities in dataset collection and refinement. This paper proposes an automatic human pose data collection system centered around a human pose estimation model, utilizing camera-based sensor fusion techniques and edge devices. The system facilitates the direct collection and real-time processing of field data at the network periphery, distributing the computational load that typically centralizes. Additionally, by directly labeling field data, it aids in constructing new training datasets.