• Korean Journal of Remote Sensing
• /
• v.38 no.6_2
• /
• pp.1633-1641
• /
• 2022

Urban population concentration and indiscriminate development are causing various environmental problems such as air pollution and heat island phenomena, and causing human resources to deteriorate the damage caused by natural disasters. Urban trees have been proposed as a solution to these urban problems, and actually play an important role, such as providing environmental improvement functions. Accordingly, quantitative measurement and analysis of individual trees in urban trees are required to understand the effect of trees on the urban environment. However, the complexity and diversity of urban trees have a problem of lowering the accuracy of single tree detection. Therefore, we conducted a study to effectively detect trees in Dongjak-gu using high-resolution aerial images that enable effective detection of tree objects and You Only Look Once Version 5 (YOLOv5), which showed excellent performance in object detection. Labeling guidelines for the construction of tree AI learning datasets were generated, and box annotation was performed on Dongjak-gu trees based on this. We tested various scale YOLOv5 models from the constructed dataset and adopted the optimal model to perform more efficient urban tree detection, resulting in significant results of mean Average Precision (mAP) 0.663.

• Journal of the Korea Society of Computer and Information
• /
• v.29 no.1
• /
• pp.51-59
• /
• 2024

In this paper, we propose a novel traffic safety system designed to reduce pedestrian traffic accidents and enhance safety on university campuses. The system involves real-time detection of vehicle speeds in designated areas and the interaction between vehicles and pedestrians at crosswalks. Utilizing the YOLOv5s model and Deep SORT method, the system performs speed measurement and object tracking within specified zones. Second, a condition-based output system is developed for crosswalk areas using the YOLOv5s object detection model to differentiate between pedestrians and vehicles. The functionality of the system was validated in real-time operation. Our system is cost-effective, allowing installation using ordinary smartphones or surveillance cameras. It is anticipated that the system, applicable not only on university campuses but also in similar problem areas, will serve as a solution to enhance safety for both vehicles and pedestrians.

• 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.

• Journal of the Korean Society of Surveying, Geodesy, Photogrammetry and Cartography
• /
• v.40 no.5
• /
• pp.381-391
• /
• 2022

This paper is intended to find one of the prevailing deep learning models that are a type of AI (Artificial Intelligence) that helps rapidly detect damaged buildings where disasters occur. The models selected are SSD-512, RetinaNet, and YOLOv3 which are widely used in object detection in recent years. These models are based on one-stage detector networks that are suitable for rapid object detection. These are often used for object detection due to their advantages in structure and high speed but not for damaged building detection in disaster management. In this study, we first trained each of the algorithms on xBD dataset that provides the post-disaster imagery with damage classification labels. Next, the three models are quantitatively evaluated with the mAP(mean Average Precision) and the FPS (Frames Per Second). The mAP of YOLOv3 is recorded at 34.39%, and the FPS reached 46. The mAP of RetinaNet recorded 36.06%, which is 1.67% higher than YOLOv3, but the FPS is one-third of YOLOv3. SSD-512 received significantly lower values than the results of YOLOv3 on two quantitative indicators. In a disaster situation, a rapid and precise investigation of damaged buildings is essential for effective disaster response. Accordingly, it is expected that the results obtained through this study can be effectively used for the rapid response in disaster management.

• Journal of Animal Science and Technology
• /
• v.63 no.6
• /
• pp.1453-1463
• /
• 2021

Feeding is the most important behavior that represents the health and welfare of weanling pigs. The early detection of feed refusal is crucial for the control of disease in the initial stages and the detection of empty feeders for adding feed in a timely manner. This paper proposes a real-time technique for the detection and recognition of small pigs using a deep-leaning-based method. The proposed model focuses on detecting pigs on a feeder in a feeding position. Conventional methods detect pigs and then classify them into different behavior gestures. In contrast, in the proposed method, these two tasks are combined into a single process to detect only feeding behavior to increase the speed of detection. Considering the significant differences between pig behaviors at different sizes, adaptive adjustments are introduced into a you-only-look-once (YOLO) model, including an angle optimization strategy between the head and body for detecting a head in a feeder. According to experimental results, this method can detect the feeding behavior of pigs and screen non-feeding positions with 95.66%, 94.22%, and 96.56% average precision (AP) at an intersection over union (IoU) threshold of 0.5 for YOLOv3, YOLOv4, and an additional layer and with the proposed activation function, respectively. Drinking behavior was detected with 86.86%, 89.16%, and 86.41% AP at a 0.5 IoU threshold for YOLOv3, YOLOv4, and the proposed activation function, respectively. In terms of detection and classification, the results of our study demonstrate that the proposed method yields higher precision and recall compared to conventional methods.

• Journal of Information Processing Systems
• /
• v.19 no.5
• /
• pp.614-630
• /
• 2023

Dead trees significantly impact forest production and the ecological environment and pose constraints to the sustainable development of forests. A lightweight YOLOv4 dead tree detection algorithm based on unmanned aerial vehicle images is proposed to address current limitations in dead tree detection that rely mainly on inefficient, unsafe and easy-to-miss manual inspections. An improved logarithmic transformation method was developed in data pre-processing to display tree features in the shadows. For the model structure, the original CSPDarkNet-53 backbone feature extraction network was replaced by MobileNetV3. Some of the standard convolutional blocks in the original extraction network were replaced by depthwise separable convolution blocks. The new ReLU6 activation function replaced the original LeakyReLU activation function to make the network more robust for low-precision computations. The K-means++ clustering method was also integrated to generate anchor boxes that are more suitable for the dataset. The experimental results show that the improved algorithm achieved an accuracy of 97.33%, higher than other methods. The detection speed of the proposed approach is higher than that of YOLOv4, improving the efficiency and accuracy of the detection process.

• Journal of Korea Multimedia Society
• /
• v.24 no.5
• /
• pp.659-666
• /
• 2021

With the continuous development of deep learning, human behavior recognition algorithms have achieved good results. However, in a multi-person recognition environment, the complex behavior environment poses a great challenge to the efficiency of recognition. To this end, this paper proposes a multi-person pose estimation model. First of all, the human detectors in the top-down framework mostly use the two-stage target detection model, which runs slow down. The single-stage YOLOv3 target detection model is used to effectively improve the running speed and the generalization of the model. Depth separable convolution, which further improves the speed of target detection and improves the model's ability to extract target proposed regions; Secondly, based on the feature pyramid network combined with context semantic information in the pose estimation model, the OHEM algorithm is used to solve difficult key point detection problems, and the accuracy of multi-person pose estimation is improved; Finally, the Euclidean distance is used to calculate the spatial distance between key points, to determine the similarity of postures in the frame, and to eliminate redundant postures.

• Journal of the Korea Society of Computer and Information
• /
• v.27 no.5
• /
• pp.47-54
• /
• 2022

Recently, with the development of deep learning technology, research on recognizing human behavior is in progress. In this paper, a study was conducted to recognize risky behaviors that may occur in a single-person household environment using deep learning technology. Due to the nature of single-person households, personal privacy protection is necessary. In this paper, we recognize human dangerous behavior in privacy protection video with Gaussian blur filters for privacy protection of individuals. The dangerous behavior recognition method uses the YOLOv5 model to detect and preprocess human object from video, and then uses it as an input value for the behavior recognition model to recognize dangerous behavior. The experiments used ResNet3D, I3D, and SlowFast models, and the experimental results show that the SlowFast model achieved the highest accuracy of 95.7% in privacy-protected video. Through this, it is possible to recognize human dangerous behavior in a single-person household environment while protecting individual privacy.

• Journal of the Korea Convergence Society
• /
• v.11 no.11
• /
• pp.9-17
• /
• 2020

In this paper, we proposed an SNS review analysis method based on deep learning for user tendency. The existing SNS review analysis method has a problem that does not reflect a variety of opinions on various interests because most are processed based on the highest weight. To solve this problem, the proposed method is to extract the user's personal tendency from the SNS review for food. It performs classification using the YOLOv3 model, and after performing a sentiment analysis through the BiLSTM model, it extracts various personal tendencies through a set algorithm. Experiments showed that the performance of Top-1 accuracy 88.61% and Top-5 90.13% for the YOLOv3 model, and 90.99% accuracy for the BiLSTM model. Also, it was shown that diversity of the individual tendencies in the SNS review classification through the heat map. In the future, it is expected to extract personal tendencies from various fields and be used for customized service or marketing.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2022.10a
• /
• pp.50-53
• /
• 2022

In order to detect an object in yolv5, a process of annotating location information on an existing image is required when learning an image. The most representative method is to draw a bounding box on an image to store location information as meta information. However, if the boundary of the object is ambiguous, it will be difficult to make a bounding box. A representative example would be to classify parts that are not fire and parts that are fire. Therefore, in this paper, images of 100 samples judged to have caught fire were learned by varying the number of boxes. The results showed better fire detection performance in the model where the bounding box was trained by annotating it with three boxes by segmenting it slightly more than annotating it with one box by holding the edge as large as possible during annotating it with one box.