• The Journal of the Korea institute of electronic communication sciences
• /
• v.17 no.2
• /
• pp.343-350
• /
• 2022

It is worth verifying the effectiveness of data integration between data with different features. This study investigated whether the data integration affects the accuracy of deep neural network (DNN), and which integration method shows the best improvement. This study used two different public datasets. One public dataset was taken in an actual farm in India. And another was taken in a laboratory environment in Korea. Leaf images were selected from two different public datasets to have five classes which includes normal and four different types of plant diseases. DNN used pre-trained VGG16 as a feature extractor and multi-layer perceptron as a classifier. Data were integrated into three different ways to be used for the training process. DNN was trained in a supervised manner via the integrated data. The trained DNN was evaluated by using a test dataset taken in an actual farm. DNN shows the best accuracy for the test dataset when DNN was first trained by images taken in the laboratory environment and then trained by images taken in the actual farm. The results show that data integration between plant images taken in a different environment helps improve the performance of deep neural networks. And the results also confirmed that independent use of plant images taken in different environments during the training process is more effective in improving the performance of DNN.

• Journal of Appropriate Technology
• /
• v.5 no.2
• /
• pp.70-81
• /
• 2019

The garment industry is one of the most labor-intensive manufacturing industries, with its sewing process relying almost entirely on manual labor. Its costs highly depend on the efficiency of this production line and thus is crucial to determine the production rate in real-time for line balancing. However, current production tracking methods are costly and make it difficult for many Small and Medium-sized Enterprises (SMEs) to implement them. As a result, their reliance on manual counting of finished products is both time consuming and prone to error, leading to high manufacturing costs and inefficiencies. In this paper, a production tracking system that uses the sewing machines' energy consumption data to track and count the total number of sewing tasks completed through Convolutional Neural Network (CNN) classifiers is proposed. This system was tested on two target sewing tasks, with a resulting maximum classification accuracy of 98.6%; all sewing tasks were detected. In the developing countries, the garment sewing industry is a very important industry, but the use of a lot of capital is very limited, such as applying expensive high technology to solve the above problem. Applied with the appropriate technology, this system is expected to be of great help to the garment industry in developing countries.

• The Journal of Korean Institute of Next Generation Computing
• /
• v.13 no.4
• /
• pp.29-39
• /
• 2017

Various electronic warfare situations drive the need to develop an integrated electronic warfare simulator that can perform electronic warfare modeling and simulation on radar threats. In this paper, we analyze the components of a simulation system to reversely model the radar threats that emit electromagnetic signals based on the parameters of the electronic information, and propose a method to gradually maintain the reverse extrapolation model of RF threats. In the experiment, we will evaluate the effectiveness of the incremental model update and also assess the integration method of reverse extrapolation models. The individual model of RF threats are constructed by using decision tree, naive Bayesian classifier, artificial neural network, and clustering algorithms through Euclidean distance and cosine similarity measurement, respectively. Experimental results show that the accuracy of reverse extrapolation models improves, while the size of the threat sample increases. In addition, we use voting, weighted voting, and the Dempster-Shafer algorithm to integrate the results of the five different models of RF threats. As a result, the final decision of reverse extrapolation through the Dempster-Shafer algorithm shows the best performance in its accuracy.

• Journal of the Korea Society of Computer and Information
• /
• v.28 no.4
• /
• pp.41-51
• /
• 2023

In this paper, we propose a technique of multi-time window feature extraction for anger detection in gait data. In the previous gait-based emotion recognition methods, the pedestrian's stride, time taken for one stride, walking speed, and forward tilt angles of the neck and thorax are calculated. Then, minimum, mean, and maximum values are calculated for the entire interval to use them as features. However, each feature does not always change uniformly over the entire interval but sometimes changes locally. Therefore, we propose a multi-time window feature extraction technique that can extract both global and local features, from long-term to short-term. In addition, we also propose an ensemble model that consists of multiple classifiers. Each classifier is trained with features extracted from different multi-time windows. To verify the effectiveness of the proposed feature extraction technique and ensemble model, a public three-dimensional gait dataset was used. The simulation results demonstrate that the proposed ensemble model achieves the best performance compared to machine learning models trained with existing feature extraction techniques for four performance evaluation metrics.

• Journal of Internet Computing and Services
• /
• v.24 no.6
• /
• pp.91-98
• /
• 2023

Recently, according to development of artificial intelligence, a wide range of industry being automatic and optimized. Also we can find out some research of using supervised learning for deteceting defect of railway in domestic rail industry. However, there are structures other than rails on the track, and the fastener is a device that binds the rail to other structures, and periodic inspections are required to prevent safety accidents. In this paper, we present a method of reducing cost for labeling using semi-supervised and transfer model trained on rail fastener data. We use Resnet50 as the backbone network pretrained on ImageNet. At first we randomly take training data from unlabeled data and then labeled that data to train model. After predict unlabeled data by trained model, we adopted a method of adding the data with the highest probability for each class to the training data by a predetermined size. Futhermore, we also conducted some experiments to investigate the influence of the number of initially labeled data. As a result of the experiment, model reaches 92% accuracy which has a performance difference of around 5% compared to supervised learning. This is expected to improve the performance of the classifier by using relatively few labels without additional labeling processes through the proposed method.

• Journal of Intelligence and Information Systems
• /
• v.29 no.4
• /
• pp.165-184
• /
• 2023

Recently, Deep learning analysis of unstructured text data using language models, such as Google's BERT and OpenAI's GPT has shown remarkable results in various applications. Most language models are used to learn generalized linguistic information from pre-training data and then update their weights for downstream tasks through a fine-tuning process. However, some concerns have been raised that privacy may be violated in the process of using these language models, i.e., data privacy may be violated when data owner provides large amounts of data to the model owner to perform fine-tuning of the language model. Conversely, when the model owner discloses the entire model to the data owner, the structure and weights of the model are disclosed, which may violate the privacy of the model. The concept of offsite tuning has been recently proposed to perform fine-tuning of language models while protecting privacy in such situations. But the study has a limitation that it does not provide a concrete way to apply the proposed methodology to text classification models. In this study, we propose a concrete method to apply offsite tuning with an additional classifier to protect the privacy of the model and data when performing multi-classification fine-tuning on Korean documents. To evaluate the performance of the proposed methodology, we conducted experiments on about 200,000 Korean documents from five major fields, ICT, electrical, electronic, mechanical, and medical, provided by AIHub, and found that the proposed plug-in model outperforms the zero-shot model and the offsite model in terms of classification accuracy.

• The Journal of the Acoustical Society of Korea
• /
• v.43 no.2
• /
• pp.225-233
• /
• 2024

Active sonar transmits sound waves to detect covertly maneuvering underwater objects and detects the signals reflected back from the target. However, in addition to the target's echo, the active sonar's received signal is mixed with seafloor, sea surface reverberation, biological noise, and other noise, making target recognition difficult. Conventional techniques for detecting signals above a threshold not only cause false detections or miss targets depending on the set threshold, but also have the problem of having to set an appropriate threshold for various underwater environments. To overcome this, research has been conducted on automatic calculation of threshold values through techniques such as Constant False Alarm Rate (CFAR) and application of advanced tracking filters and association techniques, but there are limitations in environments where a significant number of detections occur. As deep learning technology has recently developed, efforts have been made to apply it in the field of underwater target detection, but it is very difficult to acquire active sonar data for discriminator learning, so not only is the data rare, but there are only a very small number of targets and a relatively large number of non-targets. There are difficulties due to the imbalance of data. In this paper, the image of the energy distribution of the detection signal is used, and a classifier is learned in a way that takes into account the imbalance of the data to distinguish between targets and non-targets and added to the existing technique. Through the proposed technique, target misclassification was minimized and non-targets were eliminated, making target recognition easier for active sonar operators. And the effectiveness of the proposed technique was verified through sea experiment data obtained in the East Sea.

• The Transactions of the Korea Information Processing Society
• /
• v.13 no.2
• /
• pp.41-47
• /
• 2024

When there were disparities in performance between models trained in the time and frequency domains, even after conducting an ensemble, we observed that the performance of the ensemble was compromised due to imbalances in the individual model performances. Therefore, this paper proposes a leakage detection technique to enhance the accuracy of pipeline leakage detection through a step-wise learning approach that extracts features from both the time and frequency domains and integrates them. This method involves a two-step learning process. In the Stage 1, independent model training is conducted in the time and frequency domains to effectively extract crucial features from the provided data in each domain. In Stage 2, the pre-trained models were utilized by removing their respective classifiers. Subsequently, the features from both domains were fused, and a new classifier was added for retraining. The proposed transfer learning-based feature fusion technique in this paper performs model training by integrating features extracted from the time and frequency domains. This integration exploits the complementary nature of features from both domains, allowing the model to leverage diverse information. As a result, it achieved a high accuracy of 99.88%, demonstrating outstanding performance in pipeline leakage detection.

• The Journal of the Acoustical Society of Korea
• /
• v.25 no.6
• /
• pp.277-281
• /
• 2006

This paper presents a method for the automatic recognition of pitch accents over syllables. The method that we propose is based on the time-delay recursive neural network (TDRNN). which is a neural network classifier with two different representation of dynamic context: the delayed input nodes allow the representation of an explicit trajectory F0(t) along time. while the recursive nodes provide long-term context information that reflects the characteristics of pitch accentuation in spoken English. We apply the TDRNN to pitch accent recognition in two forms: in the normal TDRNN. all of the prosodic features (pitch. energy, duration) are used as an entire set in a single TDRNN. while in the distributed TDRNN. the network consists of several TDRNNs each taking a single prosodic feature as the input. The final output of the distributed TDRNN is weighted sum of the output of individual TDRNN. We used the Boston Radio News Corpus (BRNC) for the experiments on the speaker-independent pitch accent recognition. π 1e experimental results show that the distributed TDRNN exhibits an average recognition accuracy of 83.64% over both pitch events and non-events.

• Journal of Intelligence and Information Systems
• /
• v.24 no.1
• /
• pp.205-225
• /
• 2018

Convolutional Neural Network (ConvNet) is one class of the powerful Deep Neural Network that can analyze and learn hierarchies of visual features. Originally, first neural network (Neocognitron) was introduced in the 80s. At that time, the neural network was not broadly used in both industry and academic field by cause of large-scale dataset shortage and low computational power. However, after a few decades later in 2012, Krizhevsky made a breakthrough on ILSVRC-12 visual recognition competition using Convolutional Neural Network. That breakthrough revived people interest in the neural network. The success of Convolutional Neural Network is achieved with two main factors. First of them is the emergence of advanced hardware (GPUs) for sufficient parallel computation. Second is the availability of large-scale datasets such as ImageNet (ILSVRC) dataset for training. Unfortunately, many new domains are bottlenecked by these factors. For most domains, it is difficult and requires lots of effort to gather large-scale dataset to train a ConvNet. Moreover, even if we have a large-scale dataset, training ConvNet from scratch is required expensive resource and time-consuming. These two obstacles can be solved by using transfer learning. Transfer learning is a method for transferring the knowledge from a source domain to new domain. There are two major Transfer learning cases. First one is ConvNet as fixed feature extractor, and the second one is Fine-tune the ConvNet on a new dataset. In the first case, using pre-trained ConvNet (such as on ImageNet) to compute feed-forward activations of the image into the ConvNet and extract activation features from specific layers. In the second case, replacing and retraining the ConvNet classifier on the new dataset, then fine-tune the weights of the pre-trained network with the backpropagation. In this paper, we focus on using multiple ConvNet layers as a fixed feature extractor only. However, applying features with high dimensional complexity that is directly extracted from multiple ConvNet layers is still a challenging problem. We observe that features extracted from multiple ConvNet layers address the different characteristics of the image which means better representation could be obtained by finding the optimal combination of multiple ConvNet layers. Based on that observation, we propose to employ multiple ConvNet layer representations for transfer learning instead of a single ConvNet layer representation. Overall, our primary pipeline has three steps. Firstly, images from target task are given as input to ConvNet, then that image will be feed-forwarded into pre-trained AlexNet, and the activation features from three fully connected convolutional layers are extracted. Secondly, activation features of three ConvNet layers are concatenated to obtain multiple ConvNet layers representation because it will gain more information about an image. When three fully connected layer features concatenated, the occurring image representation would have 9192 (4096+4096+1000) dimension features. However, features extracted from multiple ConvNet layers are redundant and noisy since they are extracted from the same ConvNet. Thus, a third step, we will use Principal Component Analysis (PCA) to select salient features before the training phase. When salient features are obtained, the classifier can classify image more accurately, and the performance of transfer learning can be improved. To evaluate proposed method, experiments are conducted in three standard datasets (Caltech-256, VOC07, and SUN397) to compare multiple ConvNet layer representations against single ConvNet layer representation by using PCA for feature selection and dimension reduction. Our experiments demonstrated the importance of feature selection for multiple ConvNet layer representation. Moreover, our proposed approach achieved 75.6% accuracy compared to 73.9% accuracy achieved by FC7 layer on the Caltech-256 dataset, 73.1% accuracy compared to 69.2% accuracy achieved by FC8 layer on the VOC07 dataset, 52.2% accuracy compared to 48.7% accuracy achieved by FC7 layer on the SUN397 dataset. We also showed that our proposed approach achieved superior performance, 2.8%, 2.1% and 3.1% accuracy improvement on Caltech-256, VOC07, and SUN397 dataset respectively compare to existing work.