• KIPS Transactions on Software and Data Engineering
• /
• v.8 no.7
• /
• pp.311-316
• /
• 2019

Classification problem is to predict the class to which an input data belongs. One of the most popular methods to do this is training a machine learning algorithm using the given dataset. In this case, the dataset should have a well-balanced class distribution for the best performance. However, when the dataset has an imbalanced class distribution, its classification performance could be very poor. To overcome this problem, we propose an over-sampling scheme that balances the number of data by using Conditional Generative Adversarial Networks (CGAN). CGAN is a generative model developed from Generative Adversarial Networks (GAN), which can learn data characteristics and generate data that is similar to real data. Therefore, CGAN can generate data of a class which has a small number of data so that the problem induced by imbalanced class distribution can be mitigated, and classification performance can be improved. Experiments using actual collected data show that the over-sampling technique using CGAN is effective and that it is superior to existing over-sampling techniques.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.25 no.1
• /
• pp.36-43
• /
• 2021

In this research, the accuracy of YOLO v3 algorithm in object detection during AGV (Automated Guided Vehicle) operation was investigated. First of all, AGV with 2D LiDAR and stereo camera was prepared. AGV was driven along the route scanned with SLAM (Simultaneous Localization and Mapping) using 2D LiDAR while front objects were detected through stereo camera. In order to evaluate the accuracy of YOLO v3 algorithm, recall, AP (Average Precision), and mAP (mean Average Precision) of the algorithm were measured with a degree of machine learning. Experimental results show that mAP, precision, and recall are improved by 10%, 6.8%, and 16.4%, respectively, when YOLO v3 is fitted with 4000 training dataset and 500 testing dataset which were collected through online search and is trained additionally with 1200 dataset collected from the stereo camera on AGV.

• Journal of Radiation Protection and Research
• /
• v.46 no.4
• /
• pp.184-193
• /
• 2021

Background: The radionuclides released by the Fukushima Daiichi Nuclear Power Plant (FDNPP) accident 9 years ago are still being monitored by various research teams and the Japanese government. Comparison of different surveys' results could help evaluate the exposure doses and the mechanism of radiocesium behavior in the urban environment in the area. In this study, we clarified the relationship between land use and temporal changes in the ambient dose rates (air dose rates) using big data. Materials and Methods: We set a series of 1 × 1 km² meshes within the 80 km zone of the FDNPP to compare the different survey results. We then prepared an analysis dataset from all survey meshes to analyze the temporal change in the air dose rate. The selected meshes included data from all survey types (airborne, fixed point, backpack, and carborne) obtained through the all-time survey campaigns. Results and Discussion: The characteristics of each survey's results were then evaluated using this dataset, as they depended on the measurement object. The dataset analysis revealed that, for example, the results of the carborne survey were smaller than those of the other surveys because the field of view of the carborne survey was limited to paved roads. The location factor of different land uses was also evaluated considering the characteristics of the four survey methods. Nine years after the FDNPP accident, the location factor ranged from 0.26 to 0.49, while the half-life of the air dose rate ranged from 1.2 to 1.6. Conclusion: We found that the decreasing trend in the air dose rate of the FDNPP accident was similar to the results obtained after the Chernobyl accident. These parameters will be useful for the prediction of the future exposure dose at the post-accident.

• Composite Materials and Engineering
• /
• v.3 no.3
• /
• pp.201-220
• /
• 2021

In this study, potential of three machine learning techniques i.e., M5P, Support vector machines and Gaussian processes were evaluated to find the best algorithm for the prediction of flexural strength of concrete mix with steel fibre. The study comprises the comparison of results obtained from above-said techniques for given dataset. The dataset consists of 124 observations from past research studies and this dataset is randomly divided into two subsets namely training and testing datasets with (70-30)% proportion by weight. Cement, fine aggregates, coarse aggregates, water, super plasticizer/ high-range water reducer, steel fibre, fibre length and curing days were taken as input parameters whereas flexural strength of the concrete mix was taken as the output parameter. Performance of the techniques was checked by statistic evaluation parameters. Results show that the Gaussian process technique works better than other techniques with its minimum error bandwidth. Statistical analysis shows that the Gaussian process predicts better results with higher coefficient of correlation value (0.9138) and minimum mean absolute error (1.2954) and Root mean square error value (1.9672). Sensitivity analysis proves that steel fibre is the significant parameter among other parameters to predict the flexural strength of concrete mix. According to the shape of the fibre, the mixed type performs better for this data than the hooked shape of the steel fibre, which has a higher CC of 0.9649, which shows that the shape of fibers do effect the flexural strength of the concrete. However, the intricacy of the mixed fibres needs further investigations. For future mixes, the most favorable range for the increase in flexural strength of concrete mix found to be (1-3)%.

• Journal of the Society of Disaster Information
• /
• v.19 no.4
• /
• pp.968-975
• /
• 2023

Purpose: We would like to confirm that the false positive rate of flames/smoke is high when detecting fires. Propose a method and dataset to recognize and classify fire situations to reduce the false detection rate. Method: Using the video as learning data, the characteristics of the fire situation were extracted and applied to the classification model. For evaluation, the model performance of Yolov8 and Slowfast were compared and analyzed using the fire dataset conducted by the National Information Society Agency (NIA). Result: YOLO's detection performance varies sensitively depending on the influence of the background, and it was unable to properly detect fires even when the fire scale was too large or too small. Since SlowFast learns the time axis of the video, we confirmed that detects fire excellently even in situations where the shape of an atypical object cannot be clearly inferred because the surrounding area is blurry or bright. Conclusion: It was confirmed that the fire detection rate was more appropriate when using a video-based artificial intelligence detection model rather than using image data.

• The Journal of the Convergence on Culture Technology
• /
• v.10 no.2
• /
• pp.465-470
• /
• 2024

The advancement of AI-based image generation technology has resulted in the creation of various images, emphasizing the need for technology capable of accurately discerning them. The amount of generated image data is limited, and to achieve high performance with a limited dataset, this study proposes a model for discriminating generated images using transfer learning. Applying pre-trained models from the ImageNet dataset directly to the CIFAKE input dataset, we reduce training time cost followed by adding three hidden layers and one output layer to fine-tune the model. The modeling results revealed an improvement in the performance of the model when adjusting the final layer. Using transfer learning and then adjusting layers close to the output layer, small image data-related accuracy issues can be reduced and generated images can be classified.

• International Journal of Computer Science & Network Security
• /
• v.24 no.8
• /
• pp.14-20
• /
• 2024

Traffic sign recognition is an essential feature of intelligent transportation systems and Advanced Driver Assistance Systems (ADAS), which are necessary for improving road safety and advancing the development of autonomous cars. This research investigates the incorporation of the YOLOv9 model into traffic sign recognition systems, utilizing its sophisticated functionalities such as Programmable Gradient Information (PGI) and Generalized Efficient Layer Aggregation Network (GELAN) to tackle enduring difficulties in object detection. We employed a publically accessible dataset obtained from Roboflow, which consisted of 3130 images classified into five distinct categories: speed_40, speed_60, stop, green, and red. The dataset was separated into training (68%), validation (21%), and testing (12%) subsets in a methodical manner to ensure a thorough examination. Our comprehensive trials have shown that YOLOv9 obtains a mean Average Precision (mAP@0.5) of 0.959, suggesting exceptional precision and recall for the majority of traffic sign classes. However, there is still potential for improvement specifically in the red traffic sign class. An analysis was conducted on the distribution of instances among different traffic sign categories and the differences in size within the dataset. This analysis aimed to guarantee that the model would perform well in real-world circumstances. The findings validate that YOLOv9 substantially improves the precision and dependability of traffic sign identification, establishing it as a dependable option for implementation in intelligent transportation systems and ADAS. The incorporation of YOLOv9 in real-world traffic sign recognition and classification tasks demonstrates its promise in making roadways safer and more efficient.

• Convergence Security Journal
• /
• v.21 no.4
• /
• pp.69-75
• /
• 2021

Various image formats are being used when attempting to construct a malware detection model based on CNN. However, most previous studies emphasize only the detection or classification performance, and do not take into account the possible impact of image format on detection performance and resource usage. Therefore, in this paper, we analyze how the input image formats affect detection performance and resources usage when detecting android malware based on CNN. The dataset used in the experiment is the CICAndMal2017 Dataset. Subdataset extracted from the CICAndMal2017 Dataset were converted into images in four formats: BMP, JPG, PNG, and TIFF. We then trained our CNN model and measured malware detection performance and resource usage. As a result, there was no sifnificant difference between detection performance and the GPU/RAM usage, even if the image format changed. However, we found that the file size of the generated images varied by up to six times depending on the image format, and that significant differences occurred in the training time.

• Journal of Korean Society of Forest Science
• /
• v.110 no.2
• /
• pp.155-164
• /
• 2021

Many studies have been conducted on developing automatic plant identification algorithms using machine learning to various plant features, such as leaves and flowers. Unlike other plant characteristics, barks show only little change regardless of the season and are maintained for a long period. Nevertheless, barks show a complex shape with a large variation depending on the environment, and there are insufficient materials that can be utilized to train algorithms. Here, in addition to the previously published bark image dataset, BarkNet v.1.0, images of barks were collected, and a dataset consisting of 53 tree species that can be easily observed in Korea was presented. A convolutional neural network (CNN) was trained and tested on the dataset, and the factors that interfere with the model's performance were identified. For CNN architecture, VGG-16 and 19 were utilized. As a result, VGG-16 achieved 90.41% and VGG-19 achieved 92.62% accuracy. When tested on new tree images that do not exist in the original dataset but belong to the same genus or family, it was confirmed that more than 80% of cases were successfully identified as the same genus or family. Meanwhile, it was found that the model tended to misclassify when there were distracting features in the image, including leaves, mosses, and knots. In these cases, we propose that random cropping and classification by majority votes are valid for improving possible errors in training and inferences.

• Korean Journal of Remote Sensing
• /
• v.36 no.6_3
• /
• pp.1711-1720
• /
• 2020

Particulate matter (PM) that has been artificially generated during the recent of rapid industrialization and urbanization moves and disperses according to weather conditions, and adversely affects the human skin and respiratory systems. The purpose of this study is to predict the PM10 concentration in Seoul using meteorological factors as input dataset for multiple linear regression (MLR), support vector machine (SVM), and random forest (RF) models, and compared and evaluated the performance of the models. First, the PM10 concentration data obtained at 39 air quality monitoring sites (AQMS) in Seoul were divided into training and validation dataset (8:2 ratio). The nine meteorological factors (mean, maximum, and minimum temperature, precipitation, average and maximum wind speed, wind direction, yellow dust, and relative humidity), obtained by the automatic weather system (AWS), were composed to input dataset of models. The coefficients of determination (R2) between the observed PM10 concentration and that predicted by the MLR, SVM, and RF models was 0.260, 0.772, and 0.793, respectively, and the RF model best predicted the PM10 concentration. Among the AQMS used for model validation, Gwanak-gu and Gangnam-daero AQMS are relatively close to AWS, and the SVM and RF models were highly accurate according to the model validations. The Jongno-gu AQMS is relatively far from the AWS, but since PM10 concentration for the two adjacent AQMS were used for model training, both models presented high accuracy. By contrast, Yongsan-gu AQMS was relatively far from AQMS and AWS, both models performed poorly.