• Annual Conference of KIPS
• /
• 2021.05a
• /
• pp.422-425
• /
• 2021

본 논문에서는 최근 다양한 분야에서 많이 활용되고 있는 YOLO와 EAST 신경망을 이미지 속 문자열 탐지문제에 적용해보고 이들의 성능을 비교분석 해 보았다. YOLO 신경망은 v3 이전 모델까지는 이미지 속 문자영역 탐지에 낮은 성능을 보인다고 알려졌으나, 최근 출시된 YOLOv4와 YOLOv5의 경우 다양한 형태의 이미지 속에 있는 한글과 영문 문자열 탐지에 뛰어난 성능을 보여줌을 확인하고 향후 문자 인식 분야에서 많이 활용될 것으로 기대된다.

• Journal of information and communication convergence engineering
• /
• v.21 no.4
• /
• pp.329-336
• /
• 2023

This paper introduces an edge AI-based scene-specific object detection system for long-term traffic management, focusing on analyzing congestion and movement via cameras. It aims to balance fast processing and accuracy in traffic flow data analysis using edge computing. We adapt the YOLOv5 model, with four heads, to a scene-specific model that utilizes the fixed camera's scene-specific properties. This model selectively detects objects based on scale by blocking nodes, ensuring only objects of certain sizes are identified. A decision module then selects the most suitable object detector for each scene, enhancing inference speed without significant accuracy loss, as demonstrated in our experiments.

• Journal of Information Processing Systems
• /
• v.19 no.5
• /
• pp.663-672
• /
• 2023

Most vehicle detection methods have poor vehicle feature extraction performance at night, and their robustness is reduced; hence, this study proposes a night vehicle detection method based on style transfer image enhancement. First, a style transfer model is constructed using cycle generative adversarial networks (cycleGANs). The daytime data in the BDD100K dataset were converted into nighttime data to form a style dataset. The dataset was then divided using its labels. Finally, based on a YOLOv5s network, a nighttime vehicle image is detected for the reliable recognition of vehicle information in a complex environment. The experimental results of the proposed method based on the BDD100K dataset show that the transferred night vehicle images are clear and meet the requirements. The precision, recall, mAP@.5, and mAP@.5:.95 reached 0.696, 0.292, 0.761, and 0.454, respectively.

• Smart Media Journal
• /
• v.12 no.11
• /
• pp.125-133
• /
• 2023

Traditional attire has assumed a pivotal role within the contemporary fashion industry. The objective of this study is to construct a computer vision model tailored to the recognition of traditional costumes originating from five distinct countries, namely India, Korea, Japan, Tanzania, and Vietnam. Leveraging a dataset comprising 1,608 images, we proceeded to train the cutting-edge computer vision model YOLOv8. The model yielded an impressive overall mean average precision (MAP) of 96%. Notably, the Indian sari exhibited a remarkable MAP of 99%, the Tanzanian kitenge 98%, the Japanese kimono 92%, the Korean hanbok 89%, and the Vietnamese ao dai 83%. Furthermore, the model demonstrated a commendable overall box precision score of 94.7% and a recall rate of 84.3%. Within the realm of the fashion industry, this model possesses considerable utility for trend projection and the facilitation of personalized recommendation systems.

• Annual Conference of KIPS
• /
• 2024.05a
• /
• pp.760-763
• /
• 2024

실제 전투와 유사한 군사 훈련을 수행하기 위해서는 훈련 공간 확보, 악천후 극복, 실 사격 훈련, 민간인 통제 등 다양한 제약이 있다. 이러한 제약을 극복하기 위해 과학화 훈련이 도입되었으며, 현대전의 양상이 대규모 전투에서 소규모 교전으로 전환되면서 가상 훈련 시스템이 주목을 받고 있다. 가상 현실에서 적을 감지하기 위해 광선투사방식이 사용되지만, 이 방법은 인간의 시각 지각능력을 넘어서기 때문에 현실적인 훈련을 시뮬레이션 하는 데 한계가 있다. 본 논문은 가상 환경 내 가상자율군(Computer Generated Forces)이 현실적인 적 시뮬레이션을 달성하기 위하여 이미지 기반의 적 검출을 적용하여, 광선투사방식에 비해 인간 시각 지각에 더 가까운 결과를 얻었다.

• Journal of the Korean Society of Industry Convergence
• /
• v.24 no.6_2
• /
• pp.699-706
• /
• 2021

Existing license plate recognition system is used as an optical character recognition method, but a method of using deep learning has been proposed in recent studies because it has problems with image quality and Korean misrecognition. This requires a lot of data collection, but the collection of license plates is not easy to collect due to the problem of the Personal Information Protection Act, and labeling work to designate the location of individual license plates is required, but it also requires a lot of time. Therefore, in this paper, to solve this problem, five types of license plates were created using a virtual Korean license plate generation program according to the notice of the Ministry of Land, Infrastructure and Transport. And the generated license plate is synthesized in the license plate part of collectable vehicle images to construct 10,147 learning data to be used in deep learning. The learning data classifies license plates, Korean, and numbers into individual classes and learn using YOLOv5. Since the proposed method recognizes letters and numbers individually, if the font does not change, it can be recognized even if the license plate standard changes or the number of characters increases. As a result of the experiment, an accuracy of 96.82% was obtained, and it can be applied not only to the learned license plate but also to new types of license plates such as new license plates and eco-friendly license plates.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.17 no.6
• /
• pp.1249-1258
• /
• 2022

Marine debris is defined as a substance that is intentionally or inadvertently left on the shore or is introduced or discharged into the ocean, which has or is likely to have a harmful effect on the marine environments. In this study, the detection of marine debris and the analysis of the amount of change on marine debris were performed using the object detection method for an efficient method of identifying the quantity of marine debris and analyzing the amount of change. The study area is Yuho Mongdol Beach in the northeastern part of Geoje Island, and the amount of change was analyzed through images collected at 15-minute intervals for 32 days from September 12 to October 14, 2022. Marine debris detection using YOLOv5x, a one-stage object detection model, derived the performance of plastic bottles mAP 0.869 and styrofoam buoys mAP 0.862. As a result, marine debris showed a large decrease at 8-day intervals, and it was found that the quantity of Styrofoam buoys was about three times larger and the range of change was also larger.

• Journal of the Korean Society of Radiology
• /
• v.15 no.1
• /
• pp.15-20
• /
• 2021

CT is important role in the medical field, such as disease diagnosis, but the number of examination and CT images are increasing. Recently, deep learning has been actively used in the medical field, and it has been used to diagnose auxiliary disease through object detection during deep learning using medical images. The purpose of study to evaluate accuracy by detecting kidney and vertebrae during abdominal CT using object detection deep learning in YOLOv3. As a results of the study, the detection accuracy of the kidney and vertebrae was 83.00%, 82.45%, and can be used as basic data for the object detection of medical images using deep learning.

• KSCE Journal of Civil and Environmental Engineering Research
• /
• v.41 no.3
• /
• pp.317-325
• /
• 2021

Pine wilt disease first appeared in Busan in 1998; it is a serious disease that causes enormous damage to pine trees. The Korean government enacted a special law on the control of pine wilt disease in 2005, which controls and prohibits the movement of pine trees in affected areas. However, existing forecasting and control methods have physical and economic challenges in reducing pine wilt disease that occurs simultaneously and radically in mountainous terrain. In this study, the authors present the use of a deep learning object recognition and prediction method based on visual materials using an unmanned aerial vehicle (UAV) to effectively detect trees suspected of being infected with pine wilt disease. In order to observe pine wilt disease, an orthomosaic was produced using image data acquired through aerial shots. As a result, 198 damaged trees were identified, while 84 damaged trees were identified in field surveys that excluded areas with inaccessible steep slopes and cliffs. Analysis using image segmentation (SegNet) and image detection (YOLOv2) obtained a performance value of 0.57 and 0.77, respectively.

• Journal of the Korea Society of Computer and Information
• /
• v.27 no.1
• /
• pp.1-8
• /
• 2022

In this paper, we propose a fire detection system based on CCTV images using an object tracking technology with YOLOv4 model capable of real-time object detection and a DeepSORT algorithm. The fire detection model was learned from 10800 pieces of learning data and verified through 1,000 separate test sets. Subsequently, the fire detection rate in a single image and fire detection maintenance performance in the image were increased by tracking the detected fire area through the DeepSORT algorithm. It is verified that a fire detection rate for one frame in video data or single image could be detected in real time within 0.1 second. In this paper, our AI fire detection system is more stable and faster than the existing fire accident detection system.