• ETRI Journal
• /
• v.46 no.1
• /
• pp.59-70
• /
• 2024

We introduce a high-performance named entity recognition (NER) model for written and spoken language. To overcome challenges related to labeled data scarcity and domain shifts, we use transfer learning to leverage our previously developed KorBERT as the base model. We also adopt a meta-pseudo-label method using a teacher/student framework with labeled and unlabeled data. Our model presents two modifications. First, the student model is updated with an average loss from both human- and pseudo-labeled data. Second, the influence of noisy pseudo-labeled data is mitigated by considering feedback scores and updating the teacher model only when below a threshold (0.0005). We achieve the target NER performance in the spoken language domain and improve that in the written language domain by proposing a straightforward rollback method that reverts to the best model based on scarce human-labeled data. Further improvement is achieved by adjusting the label vector weights in the named entity dictionary.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.18 no.3
• /
• pp.738-754
• /
• 2024

Side-channel analysis (SCA) is a cryptanalytic technique that exploits physical leakages, such as power consumption or electromagnetic emanations, from cryptographic devices to extract secret keys used in cryptographic algorithms. Recent studies have shown that training SCA models with semi-supervised learning can effectively overcome the problem of few labeled power traces. However, the process of training SCA models using semi-supervised learning generates many pseudo-labels. The performance of the SCA model can be reduced by some of these pseudo-labels. To solve this issue, we propose the HWFilter method to improve semi-supervised SCA. This method uses a Hamming Weight Pseudo-label Filter (HWPF) to filter the pseudo-labels generated by the semi-supervised SCA model, which enhances the model's performance. Furthermore, we introduce a normal distribution method for constructing the HWPF. In the normal distribution method, the Hamming weights (HWs) of power traces can be obtained from the normal distribution of power points. These HWs are filtered and combined into a HWPF. The HWFilter was tested using the ASCADv1 database and the AES_HD dataset. The experimental results demonstrate that the HWFilter method can significantly enhance the performance of semi-supervised SCA models. In the ASCADv1 database, the model with HWFilter requires only 33 power traces to recover the key. In the AES_HD dataset, the model with HWFilter outperforms the current best semi-supervised SCA model by 12%.

• Journal of Institute of Control, Robotics and Systems
• /
• v.18 no.1
• /
• pp.35-41
• /
• 2012

This paper proposes a moving object detection algorithm for active camera system that can be applied to mobile robot and intelligent surveillance system. Most of moving object detection algorithms based on a stationary camera system. These algorithms used fixed surveillance system that does not consider the motion of the background or robot tracking system that track pre-learned object. Unlike the stationary camera system, the active camera system has a problem that is difficult to extract the moving object due to the error occurred by the movement of camera. In order to overcome this problem, the motion of the camera was compensated by using SURF and Pseudo Perspective model, and then the moving object is extracted efficiently using stochastic Label Cluster transport model. This method is possible to detect moving object because that minimizes effect of the background movement. Our approach proves robust and effective in terms of moving object detection in active camera system.

• Journal of Dental Rehabilitation and Applied Science
• /
• v.38 no.4
• /
• pp.196-203
• /
• 2022

Purpose: This study aimed to evaluate the accuracy and clinical usability of an identification model using deep learning for 79 dental implant types. Materials and Methods: A total of 45396 implant fixture images were collected through panoramic radiographs of patients who received implant treatment from 2001 to 2020 at 30 dental clinics. The collected implant images were 79 types from 18 manufacturers. EfficientNet and Meta Pseudo Labels algorithms were used. For EfficientNet, EfficientNet-B0 and EfficientNet-B4 were used as submodels. For Meta Pseudo Labels, two models were applied according to the widen factor. Top 1 accuracy was measured for EfficientNet and top 1 and top 5 accuracy for Meta Pseudo Labels were measured. Results: EfficientNet-B0 and EfficientNet-B4 showed top 1 accuracy of 89.4. Meta Pseudo Labels 1 showed top 1 accuracy of 87.96, and Meta pseudo labels 2 with increased widen factor showed 88.35. In Top5 Accuracy, the score of Meta Pseudo Labels 1 was 97.90, which was 0.11% higher than 97.79 of Meta Pseudo Labels 2. Conclusion: All four deep learning algorithms used for implant identification in this study showed close to 90% accuracy. In order to increase the clinical applicability of deep learning for implant identification, it will be necessary to collect a wider amount of data and develop a fine-tuned algorithm for implant identification.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.24 no.10
• /
• pp.1271-1279
• /
• 2020

Text region detection is a technology that detects text area in handwriting or printed documents. The detected text areas are digitized through a recognition step, which is used in various fields depending on the purpose of use. However, the detection result of the small text unit is not suitable for the industrial field. In addition, the border of tables in the document that it causes miss-detected results, which has an adverse effect on the recognition step. To solve the issues, we propose a method for detecting text region using the border information of the table. In order to utilize the border information of the table, the proposed method adjusts the flow of two decoders. Experimentally, we show improved performance using the table border pseudo label based on weak supervised learning.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.7 no.3
• /
• pp.532-543
• /
• 2003

In this paper, we proposed structure of a reversible discrete-time cellular neural network (DTCNN) for labeling digital images to protect copylight. First, we present the concept and the structure of reversible DTCNN, which can be used to generate 2D binary pseudo-random images sequences. We presented some, output examples of different kinds of reversible DTCNNs to show their complex behaviors. Then both the original image and the copyright label, which is often another binary image, are used to generate a binary random key image. The key image is then used to scramble the original image. Since the reversibility of a reversible DTCNN, the same reversible DTCNN can recover the copyright label from a labeled image. Due to the high speed of a DTCNN chip, our method can be used to label image sequences, e.g., video sequences, in real time. Computer simulation results are presented.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.24 no.1
• /
• pp.83-88
• /
• 2024

In this research, we propose a Semi-Supervised learning based railroad surface defect detection method. The Resnet50 model, pretrained on ImageNet, was employed for the training. Data without labels are randomly selected, and then labeled to train the ResNet50 model. The trained model is used to predict the results of the remaining unlabeled training data. The predicted values exceeding a certain threshold are selected, sorted in descending order, and added to the training data. Pseudo-labeling is performed based on the class with the highest probability during this process. An experiment was conducted to assess the overall class classification performance based on the initial number of labeled data. The results showed an accuracy of 98% at best with less than 10% labeled training data compared to the overall training data.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2021.05a
• /
• pp.607-610
• /
• 2021

When learning a model in supervised learning, input data and the label of the data are required. However, labeling is high cost task and if automated, there is no guarantee that the label will always be correct. In the case of supervised learning in such a noisy labels environment, the accuracy of the model increases at the initial stage of learning, but decrease significantly after a certain period of time. There are various methods to solve the noisy label problem. But in most cases, the probability predicted by the model is used as the pseudo label. So, we proposed a method to predict the true label more quickly by refining the probabilities predicted by the model. Result of experiments on the same environment and dataset, it was confirmed that the performance improved and converged faster. Through this, it can be applied to methods that use the probability distribution predicted by the model among existing studies. And it is possible to reduce the time required for learning because it can converge faster in the same environment.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2023.05a
• /
• pp.510-512
• /
• 2023

In order to enhance a model's capability for detecting facial expressions, this research suggests a pipeline that makes use of the GradCAM component. The patching module and the pseudo-labeling module make up the pipeline. The patching component takes the original face image and divides it into four equal parts. These parts are then each input into a 2Dconvolutional layer to produce a feature vector. Each picture segment is assigned a weight token using GradCAM in the pseudo-labeling module, and this token is then merged with the feature vector using principal component analysis. A convolutional neural network based on transfer learning technique is then utilized to extract the deep features. This technique applied on a public dataset MMI and achieved a validation accuracy of 96.06% which is showing the effectiveness of our method.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2022.11a
• /
• pp.628-629
• /
• 2022

딥러닝의 높은 예측 정확도를 위해서는 많은 양의 학습 데이터가 필요하다. 그러나 실세계에서 많은 양의 레이블이 붙은 데이터를 구하는 것은 어렵고 많은 비용이 든다. 때문에 레이블이 없이도 양질의 표현 학습이 가능한 자기지도학습이 각광을 받고 있다. 와서스타인 거리는 생성모델에도 쓰이지만 의사 레이블 (pseudo label) 을 만들어 레이블이 없는 데이터들을 분류 하는데도 좋은 성능을 보이고 있다. 따라서. 본 연구는 와서스타인 거리를 자기지도학습에 접목시키는 방법을 제안한다. 실험을 통해 연구의 가능성을 보인다.