• Annual Conference on Human and Language Technology
• /
• 2022.10a
• /
• pp.88-91
• /
• 2022

디지털 휴먼, 민원 상담, ARS 등 칫챗의 활용과 수요가 증가함에 따라 칫챗의 성능 향상을 위한 다양한 연구가 진행되고 있다. 특히, 오토 인코더(Auto-encoder) 기반의 생성 모델(Generative Model)은 높은 성능을 보이며 지속적인 연구가 이루어지고 있으나, 이전 대화들에 대한 충분한 문맥 정보의 반영이 어렵고 문법적으로 부적절한 답변을 생성하는 문제가 있다. 이를 개선하기 위해 검색 기반의 생성 모델과 관련된 연구가 진행되고 있으나, 현재 시점의 문장이 유사해도 이전 문장들에 따라 의도와 답변이 달라지는 멀티턴 대화 특징을 반영하여 대화를 검색하는 연구가 부족하다. 본 논문에서는 이와 같은 멀티턴 대화의 특징이 고려된 검색 방법을 제안하고 검색된 레퍼런스(준정답 문장)를 멀티턴 대화와 함께 생성 모델의 입력으로 활용하여 학습시키는 방안을 제안한다. 제안 방안으로 학습된 발화 생성 모델은 기존 모델과 비교 평가를 수행하며 Rouge-1 스코어에서 13.11점, Rouge-2 스코어에서 10.09점 Rouge-L 스코어에서 13.2점 향상된 성능을 보였고 이를 통해 제안 방안의 우수성을 입증하였다.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2021.11a
• /
• pp.827-829
• /
• 2021

현대사회의 시설물 중 다수가 콘크리트를 사용하여 건설되었고, 재료적 성질로 인해 균열, 박락, 백태 등의 손상이 발생하고 있고 시설물 관리가 요구되고 있다. 하지만, 현재 시설물 관리는 사람의 육안 점검을 정기적으로 수행하고 있으나, 높은 시설물이나 맨눈으로 확인할 수 없는 시설물의 경우 관리가 어렵다. 이에 본 논문에서는 다양한 영상장비를 활용해 시설물의 이미지에서 균열을 분류하는 알고리즘을 제안한다. 균열 분류 알고리즘은 산업 이상 감지 데이터 세트인 MVTec AD 데이터 세트를 사전 학습하고 L2 auto-encoder를 사용하여 균열을 분류한다. MVTec AD 데이터 세트를 사전학습시킴으로써 균열, 박락, 백태 등의 특징을 학습시킬 수 있을 것으로 기대한다.

• International Journal of Internet, Broadcasting and Communication
• /
• v.15 no.4
• /
• pp.74-83
• /
• 2023

In this paper, we propose an image compression scheme which uses a demosaicking network and a channel-wise decoder in the decoding network. For the demosaicing network, we use as the input a colored mosaiced pattern rather than the well-known Bayer pattern. The use of a colored mosaiced pattern results in the mosaiced image containing a greater amount of information pertaining to the original image. Therefore, it contributes to result in a better color reconstruction. The channel-wise decoder is composed of multiple decoders where each decoder is responsible for each channel in the color image, i.e., the R, G, and B channels. The encoder and decoder are both implemented by wavelet based auto-encoders for better performance. Experimental results verify that the separated channel-wise decoders and the colored mosaic pattern produce a better reconstructed color image than a single decoder. When combining the colored CFA with the multi-decoder, the PSNR metric exhibits an increase of over 2dB for three-times compression and approximately 0.6dB for twelve-times compression compared to the Bayer CFA with a single decoder. Therefore, the compression rate is also increased with the proposed method than with the method using a single decoder on the Bayer patterned mosaic image.

• Journal of the Korea Institute of Military Science and Technology
• /
• v.26 no.5
• /
• pp.422-430
• /
• 2023

To build an automated system detecting toxic chemicals from Raman spectra, we have to obtain sufficient data of toxic chemicals. However, it usually costs high to gather Raman spectra of toxic chemicals in diverse situations. Tackling this problem, we develop methods to generate synthetic Raman spectra of DMMP and 2-CEES without actual experiments. First, we propose certain mathematical transforms to augment few original Raman spectra. Then, we train deep generative models to generate more realistic and diverse data. Analyzing synthetic Raman spectra of toxic chemicals generated by our methods through visualization, we qualitatively verify that the data are sufficiently similar to original data and diverse. For conclusion, we obtain a synthetic dataset of DMMP and 2-CEES with the proposed algorithm.

• Journal of Broadcast Engineering
• /
• v.27 no.1
• /
• pp.31-43
• /
• 2022

Traditional video compression has developed so far based on hybrid compression methods through motion prediction, residual coding, and quantization. With the rapid development of technology through artificial neural networks in recent years, research on image compression and video compression based on artificial neural networks is also progressing rapidly, showing competitiveness compared to the performance of traditional video compression codecs. In this paper, a new method capable of improving the performance of such an artificial neural network-based video compression model is presented. Basically, we take the rate-distortion optimization method using the auto-encoder and entropy model adopted by the existing learned video compression model and shifts some components of the latent information that are difficult for entropy model to estimate when transmitting compressed latent representation to the decoder side from the encoder side, and finally compensates the distortion of lost information. In this way, the existing neural network based video compression framework, MFVC (Motion Free Video Compression) is improved and the BDBR (Bjøntegaard Delta-Rate) calculated based on H.264 is nearly twice the amount of bits (-27%) of MFVC (-14%). The proposed method has the advantage of being widely applicable to neural network based image or video compression technologies, not only to MFVC, but also to models using latent information and entropy model.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.26 no.6
• /
• pp.834-841
• /
• 2022

Recently, applying artificial intelligence technologies in various fields of production has drawn an upsurge of research interest due to the increase for smart factory and artificial intelligence technologies. A great deal of effort is being made to introduce artificial intelligence algorithms into the defect detection task. Particularly, detection of defects on the surface of metal has a higher level of research interest compared to other materials (wood, plastics, fibers, etc.). In this paper, we compare and analyze the speed and performance of defect classification by combining machine learning techniques (Support Vector Machine, Softmax Regression, Decision Tree) with dimensionality reduction algorithms (Principal Component Analysis, AutoEncoders) and two convolutional neural networks (proposed method, ResNet). To validate and compare the performance and speed of the algorithms, we have adopted two datasets ((i) public dataset, (ii) actual dataset), and on the basis of the results, the most efficient algorithm is determined.

• KIPS Transactions on Software and Data Engineering
• /
• v.10 no.12
• /
• pp.579-586
• /
• 2021

Time series forecasting refers to predicting future time information based on past time information. Accurately predicting future information is crucial because it is used for establishing strategies or making policy decisions in various fields. Recently, a transformer model has been mainly studied for a time series prediction model. However, the existing transformer model has a limitation in that it has an auto-regressive structure in which the output result is input again when the prediction sequence is output. This limitation causes a problem in that accuracy is lowered when predicting a distant time point. This paper proposes a sequential decoding model focusing on the style transformation technique to handle these problems and make more precise time series forecasting. The proposed model has a structure in which the contents of past data are extracted from the transformer-encoder and reflected in the style-based decoder to generate the predictive sequence. Unlike the decoder structure of the conventional auto-regressive transformer, this structure has the advantage of being able to more accurately predict information from a distant view because the prediction sequence is output all at once. As a result of conducting a prediction experiment with various time series datasets with different data characteristics, it was shown that the model presented in this paper has better prediction accuracy than other existing time series prediction models.

• Journal of the Korea Institute of Information Security & Cryptology
• /
• v.30 no.4
• /
• pp.593-603
• /
• 2020

The advent of malicious code has increased exponentially due to the spread of malicious code generation tools in accordance with the development of the network, but there is a limit to the response through existing malicious code detection methods. According to this situation, a machine learning-based malicious code detection method is evolving, and in this paper, the feature of data is extracted from the PE header for machine-learning-based malicious code detection, and then it is used to automate the malware through autoencoder. Research on how to extract the indicated features and feature importance. In this paper, 549 features composed of information such as DLL/API that can be identified from PE files that are commonly used in malware analysis are extracted, and autoencoder is used through the extracted features to improve the performance of malware detection in machine learning. It was proved to be successful in providing excellent accuracy and reducing the processing time by 2 times by effectively extracting the features of the data by compressively storing the data. The test results have been shown to be useful for classifying malware groups, and in the future, a classifier such as SVM will be introduced to continue research for more accurate malware detection.

• KIPS Transactions on Software and Data Engineering
• /
• v.6 no.12
• /
• pp.573-580
• /
• 2017

Deep Neural Network (DNN) is a large layered neural network which is consisted of a number of layers of non-linear units. Deep Learning which represented as DNN has been applied very successfully in various applications. However, many issues in DNN have been identified through past researches. Among these issues, generalization is the most well-known problem. A Recent study, Dropout, successfully addressed this problem. Also, Dropout plays a role as noise, and so it helps to learn robust feature during learning in DNN such as Denoising AutoEncoder. However, because of a large computations required in Dropout, training takes a lot of time. Since Dropout keeps changing an inter-layer representation during the training session, the learning rates should be small, which makes training time longer. In this paper, using mutation operation, we reduce computation and improve generalization performance compared with Dropout. Also, we experimented proposed method to compare with Dropout method and showed that our method is superior to the Dropout one.

• The Journal of the Acoustical Society of Korea
• /
• v.39 no.5
• /
• pp.447-453
• /
• 2020

In this paper, we propose a system to extract effective speaker representations from a speech signal using a deep learning method. Based on the fact that speech signal contains identity unrelated information such as text content, emotion, background noise, and so on, we perform a training such that the extracted features only represent speaker-related information but do not represent speaker-unrelated information. Specifically, we propose an auto-encoder based disentanglement method that outputs both speaker-related and speaker-unrelated embeddings using effective loss functions. To further improve the reconstruction performance in the decoding process, we also introduce a discriminator popularly used in Generative Adversarial Network (GAN) structure. Since improving the decoding capability is helpful for preserving speaker information and disentanglement, it results in the improvement of speaker verification performance. Experimental results demonstrate the effectiveness of our proposed method by improving Equal Error Rate (EER) on benchmark dataset, Voxceleb1.