• Journal of the Korea Institute of Military Science and Technology
• /
• v.27 no.2
• /
• pp.256-264
• /
• 2024

In the last 10 years, AI has made rapid progress, and image classification, in particular, are showing excellent performance based on deep learning. Nevertheless, due to the nature of deep learning represented by a black box, it is difficult to actually use it in critical decision-making situations such as national defense, autonomous driving, medical care, and finance due to the lack of explainability of judgement results. In order to overcome these limitations, in this study, a model description algorithm capable of local interpretation was applied to the inception network-derived AI to analyze what grounds they made when classifying national defense data. Specifically, we conduct a comparative analysis of explainability based on confidence values by performing LIME analysis from the Inception v2_resnet model and verify the similarity between human interpretations and LIME explanations. Furthermore, by comparing the LIME explanation results through the Top1 output results for Inception v3, Inception v2_resnet, and Xception models, we confirm the feasibility of comparing the efficiency and availability of deep learning networks using XAI.

• Journal of Broadcast Engineering
• /
• v.27 no.1
• /
• pp.133-145
• /
• 2022

Compared to the continuously increasing dog population and industry size in Korea, systematic analysis of related data and research on breed classification methods are very insufficient. In this paper, an automatic breed classification method is proposed using deep learning technology for 14 major dog breeds domestically raised. To do this, dog images are collected for deep learning training and a dataset is built, and a breed classification algorithm is created by performing transfer learning based on VGG-16 and Resnet-34 as backbone networks. In order to check the transfer learning effect of the two models on dog images, we compared the use of pre-trained weights and the experiment of updating the weights. When fine tuning was performed based on VGG-16 backbone network, in the final model, the accuracy of Top 1 was about 89% and that of Top 3 was about 94%, respectively. The domestic dog breed classification method and data construction proposed in this paper have the potential to be used for various application purposes, such as classification of abandoned and lost dog breeds in animal protection centers or utilization in pet-feed industry.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.20 no.1
• /
• pp.219-223
• /
• 2020

To guarantee AI model's prominent recognition rate and recognition precision, obtaining the large number of data is essential. In this paper, we propose transfer learning-based object detection algorithm for maintaining outstanding performance even when the volume of training data is small. Also, we proposed a tranfer learning network combining Resnet-50 and YOLO(You Only Look Once) network. The transfer learning network uses the Leeds Sports Pose dataset to train the network that detects the person who occupies the largest part of each images. Simulation results yield to detection rate as 84% and detection precision as 97%.

• Journal of IKEEE
• /
• v.26 no.4
• /
• pp.647-652
• /
• 2022

In this paper, we propose a prediction system for skin pore labeling based on a CNN(Convolution Neural Network) model, where a data set is constructed by processing skin images taken by users, and a pore feature image is generated by the proposed image processing algorithm. The skin image data set was labeled for pore characteristics based on the visual classification criteria of skin beauty experts. The proposed image processing algorithm was applied to generate pore feature images from skin images and to train a CNN model that predicts pore feature ratings. The prediction results with pore features by the proposed CNN model is similar to experts visual classification results, where less learning time and higher prediction results were obtained than the results by the comparison model (Resnet-50). In this paper, we describe the proposed image processing algorithm and CNN model, the results of the prediction system and future research plans.

• Animal Bioscience
• /
• v.36 no.6
• /
• pp.980-989
• /
• 2023

Objective: Iris pattern recognition system is well developed and practiced in human, however, there is a scarcity of information on application of iris recognition system in animals at the field conditions where the major challenge is to capture a high-quality iris image from a constantly moving non-cooperative animal even when restrained properly. The aim of the study was to validate and identify Black Bengal goat biometrically to improve animal management in its traceability system. Methods: Forty-nine healthy, disease free, 3 months±6 days old female Black Bengal goats were randomly selected at the farmer's field. Eye images were captured from the left eye of an individual goat at 3, 6, 9, and 12 months of age using a specialized camera made for human iris scanning. iGoat software was used for matching the same individual goats at 3, 6, 9, and 12 months of ages. Resnet152V2 deep learning algorithm was further applied on same image sets to predict matching percentages using only captured eye images without extracting their iris features. Results: The matching threshold computed within and between goats was 55%. The accuracies of template matching of goats at 3, 6, 9, and 12 months of ages were recorded as 81.63%, 90.24%, 44.44%, and 16.66%, respectively. As the accuracies of matching the goats at 9 and 12 months of ages were low and below the minimum threshold matching percentage, this process of iris pattern matching was not acceptable. The validation accuracies of resnet152V2 deep learning model were found 82.49%, 92.68%, 77.17%, and 87.76% for identification of goat at 3, 6, 9, and 12 months of ages, respectively after training the model. Conclusion: This study strongly supported that deep learning method using eye images could be used as a signature for biometric identification of an individual goat.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.23 no.6
• /
• pp.1-7
• /
• 2023

This paper developed a training model to classify normal roads and flooded roads using artificial intelligence technology. We expanded the diversity of learning data using various data augmentation techniques and implemented a model that shows good performance in various environments. Transfer learning was performed using the CNN-based Resnet152v2 model as a pre-learning model. During the model learning process, the performance of the final model was improved through various parameter tuning and optimization processes. Learning was implemented in Python using Google Colab NVIDIA Tesla T4 GPU, and the test results showed that flooding situations were detected with very high accuracy in the test dataset.

• 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
• /
• 2018.10a
• /
• pp.364-368
• /
• 2018

In this paper, we use Faster R-CNN that is one of object detection algorithm and OpenCV that purposes computer vision, to implement the system that can detect danger when a vehicle attempts to change lanes into its own lane in videos of highway, national road, general road and etc. Also, the performance of implemented system is evaluated to prove that the performance is not bad.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2019.05a
• /
• pp.525-528
• /
• 2019

Atmospheric particle removal is a challenging task and attacks wide interests in computer vision filed. In this paper, we proposed a single image snow removal framework based on deep residual networks. According to the fact that there are various snow sizes in a snow image, the inception module which consists of different filter kernels was adopted to extract multiple resolution features of the input snow image. Except the traditional mean square error loss, the perceptual loss and total variation loss were employed to generate more clean images. Experimental results on synthetic and realistic snow images indicated that the proposed method achieves superior performance in respect of visual perception and objective evaluation.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• fall
• /
• pp.91-94
• /
• 2021

As deep learning technologies becoming developed, realistic fake videos synthesized by deep learning models called "Deepfake" videos became even more difficult to distinguish from original videos. As fake news or Deepfake blackmailing are causing confusion and serious problems, this paper suggests a novel model detecting Deepfake videos. We chose Residual Convolutional Neural Network (Resnet50) as an extraction model and Long Short-Term Memory (LSTM) which is a form of Recurrent Neural Network (RNN) as a classification model. We adopted cosine similarity with hinge loss to train our extraction model in embedding the features of Deepfake and original video. The result in this paper demonstrates that temporal features in the videos are essential for detecting Deepfake videos.