• 한국전기전자재료학회논문지
• /
• 제37권3호
• /
• pp.261-273
• /
• 2024

This paper is an experimental study on the improvement of smoke and flame detection from different materials with YOLO. For the study, images of fires occurring in various materials were collected through an open dataset, and experiments were conducted by changing the main factors affecting the performance of the fire object detection model, such as the bounding box, polygon, and data augmentation of the collected image open dataset during data preprocessing. To evaluate the model performance, we calculated the values of precision, recall, F1Score, mAP, and FPS for each condition, and compared the performance of each model based on these values. We also analyzed the changes in model performance due to the data preprocessing method to derive the conditions that have the greatest impact on improving the performance of the fire object detection model. The experimental results showed that for the fire object detection model using the YOLOv5s6.0 model, data augmentation that can change the color of the flame, such as saturation, brightness, and exposure, is most effective in improving the performance of the fire object detection model. The real-time fire object detection model developed in this study can be applied to equipment such as existing CCTV, and it is believed that it can contribute to minimizing fire damage by enabling early detection of fires occurring in various materials.

• ETRI Journal
• /
• 제42권3호
• /
• pp.411-419
• /
• 2020

In real-world intelligent transportation systems, accuracy in vehicle license plate detection and recognition is considered quite critical. Many algorithms have been proposed for still images, but their accuracy on actual videos is not satisfactory. This stems from several problematic conditions in videos, such as vehicle motion blur, variety in viewpoints, outliers, and the lack of publicly available video datasets. In this study, we focus on these challenges and propose a license plate detection and recognition scheme for videos based on a temporal matching prior network. Specifically, to improve the robustness of detection and recognition accuracy in the presence of motion blur and outliers, forward and bidirectional matching priors between consecutive frames are properly combined with layer structures specifically designed for plate detection. We also built our own video dataset for the deep training of the proposed network. During network training, we perform data augmentation based on image rotation to increase robustness regarding the various viewpoints in videos.

• 전력전자학회:학술대회논문집
• /
• 전력전자학회 2018년도 전력전자학술대회
• /
• pp.275-277
• /
• 2018

Recently, the importance of an early response has been emphasized due to the large fire. The most efficient method of extinguishing a large fire is early response to a small flame. To implement this solution, we propose a fire detection mechanism based on a deep learning artificial intelligence. In this study, a small amount of data sets is manipulated by an image augmentation technique using rotating, tilting, blurring, and distorting effects in order to increase the number of the data sets by 5 times, and we study the flame detection algorithm using faster R-CNN.

• Maxillofacial Plastic and Reconstructive Surgery
• /
• 제22권1호
• /
• pp.105-109
• /
• 2000

The zygoma is protruded on both sides of the midface and plays an important part in determining the impression of face. There are much different esthetic consideration of zygomatic prominence between the oriental and western. Because slight prominent zygoma is thought to be esthetic in the western, there has been done many augmentation therapy. But, in the oriental, prominent zygoma, especially in the female, is thought to be aggresive and manlike image, there has been done many zygoma reduction therapy. There are two methods to reduce the prominent zygoma, one is shaving method and the other is zygoma repositioning with ostectomy. The approach to zygoma and zygomatic arch without surgical morbidity is very difficult, so this problem has been difficult subject to oral & maxillofacial surgeon in a long period. We report the treatment case of facial asymmetry with the unilateral zygoma reduction & genioplasty using only intraoral approach with the retrospective study.

• 한국컴퓨터정보학회:학술대회논문집
• /
• 한국컴퓨터정보학회 2019년도 제59차 동계학술대회논문집 27권1호
• /
• pp.123-124
• /
• 2019

This paper proposes an improved deep learning method based on small data sets for animal image classification. Firstly, we use a CNN to build a training model for small data sets, and use data augmentation to expand the data samples of the training set. Secondly, using the pre-trained network on large-scale datasets, such as VGG16, the bottleneck features in the small dataset are extracted and to be stored in two NumPy files as new training datasets and test datasets. Finally, training a fully connected network with the new datasets. In this paper, we use Kaggle famous Dogs vs Cats dataset as the experimental dataset, which is a two-category classification dataset.

• 전기학회논문지
• /
• 제68권2호
• /
• pp.359-363
• /
• 2019

As commercial drones have been widely used, concerns for collision accidents with people and invading secured properties are emerging. The detection of drone is a challenging problem. The deep learning based object detection techniques for detecting drones have been applied, but limited to the specific cases such as detection of drones from bird and/or background. We have tried not only detection of drones, but classification of different drones with an end-to-end model. YOLOv2 is used as an object detection model. In order to supplement insufficient data by shooting drones, data augmentation from collected images is executed. Also transfer learning from ImageNet for YOLOv2 darknet framework is performed. The experimental results for drone detection with average IoU and recall are compared and analysed.

• 한국정보처리학회:학술대회논문집
• /
• 한국정보처리학회 2020년도 춘계학술발표대회
• /
• pp.544-546
• /
• 2020

의료 데이터를 이용하여 인공지능 기계학습 연구를 수행할 때 자주 마주하는 문제는 데이터 불균형, 데이터 부족 등이며 특히 정제된 충분한 데이터를 구하기 힘들다는 것이 큰 문제이다. 본 연구에서는 이를 해결하기 위해 GAN(Generative Adversarial Network) 기반 고해상도 의료 영상을 생성하는 프레임워크를 개발하고자 한다. 각 해상도 마다 Scale 의 Gradient 를 동시에 학습하여 빠르게 고해상도 이미지를 생성해낼 수 있도록 했다. 고해상도 이미지를 생성하는 Neural Network 를 고안하였으며, PGGAN, Style-GAN 과의 성능 비교를 통해 제안된 모델이 양질의 고해상도 의료영상 이미지를 더 빠르게 생성할 수 있음을 확인하였다. 이를 통해 인공지능 기계학습 연구에 있어서 의료 영상의 데이터 부족, 데이터 불균형 문제를 해결할 수 있는 Data augmentation 이나, Anomaly detection 등의 연구에 적용할 수 있다.

• 한국방송∙미디어공학회:학술대회논문집
• /
• 한국방송∙미디어공학회 2022년도 하계학술대회
• /
• pp.217-220
• /
• 2022

본 논문에서는 CCTV 영상 화질을 향상하고 해상도를 높이기 위해 딥 러닝(Deep Learning)을 이용하여 잡음 제거(Denoising) 와 초해상도(Super-resolution) 작업을 수행한다. 데이터 증강(Data Augmentation)을 통한 초해상도 성능 향상을 위해서 잡음 제거 네트워크의 출력 영상을 초해상도 네트워크의 입력으로 사용하는 순차적 작업을 사용한다. 또한 딥 러닝을 이용한 영상처리에서 발생하는 평균 밝기 오차 문제를 해결하기 위한 손실함수(Loss Function)와 두 가지 이상의 순차적인 딥 러닝 작업에서 발생하는 문제점을 극복하기 위한 손실함수를 제안한다. 제안하는 손실함수는 네트워크의 출력 영상과 타겟 영상의 밝기 오차를 줄이는 것이 가능하고, 순차적 작업에서 보다 정확한 모델 성능 판단이 가능하다.

• 한국컴퓨터정보학회:학술대회논문집
• /
• 한국컴퓨터정보학회 2024년도 제69차 동계학술대회논문집 32권1호
• /
• pp.99-100
• /
• 2024

다양한 변종들의 존재와 잘 알려지지 않은 취약점을 이용한 공격은 악성코드 수집을 어렵게 하는 요인들이다. 부족한 악성코드 수를 보완하고자 생성 모델을 활용한 이미지 기반의 악성코드 데이터를 증강한 연구들도 존재하였다. 하지만 생성 모델이 실제 악성코드를 생성할 수 있는지에 대한 분석은 진행되지 않았다. 본 연구는 VGG-11 모델을 활용해 실제 악성코드와 생성된 악성코드 이미지의 이진 분류하였다. 실험 결과 VGG-11 모델은 99.9%의 정확도로 두 영상을 다르게 판단한다

• ETRI Journal
• /
• 제46권3호
• /
• pp.526-537
• /
• 2024

Monitoring fetal growth in utero is crucial to anomaly diagnosis. However, current computer-vision models struggle to accurately assess the key metrics (i.e., head circumference and occipitofrontal and biparietal diameters) from ultrasound images, largely owing to a lack of training data. Mitigation usually entails image augmentation (e.g., flipping, rotating, scaling, and translating). Nevertheless, the accuracy of our task remains insufficient. Hence, we offer a U-Net fetal head measurement tool that leverages a hybrid Dice and binary cross-entropy loss to compute the similarity between actual and predicted segmented regions. Ellipse-fitted two-dimensional ultrasound images acquired from the HC18 dataset are input, and their lower feature layers are reused for efficiency. During regression, a novel region of interest pooling layer extracts elliptical feature maps, and during segmentation, feature pyramids fuse field-layer data with a new scale attention method to reduce noise. Performance is measured by Dice similarity, mean pixel accuracy, and mean intersection-over-union, giving 97.90%, 99.18%, and 97.81% scores, respectively, which match or outperform the best U-Net models.