• Journal of information and communication convergence engineering
• /
• v.19 no.2
• /
• pp.79-83
• /
• 2021

This paper proposes a novel image classification method based on few-shot learning, which is mainly used to solve model overfitting and non-convergence in image classification tasks of small datasets and improve the accuracy of classification. This method uses model structure optimization to extend the basic convolutional neural network (CNN) model and extracts more image features by adding convolutional layers, thereby improving the classification accuracy. We incorporated certain measures to improve the performance of the model. First, we used general methods such as setting a lower learning rate and shuffling to promote the rapid convergence of the model. Second, we used the data expansion technology to preprocess small datasets to increase the number of training data sets and suppress over-fitting. We applied the model to 10 monkey species and achieved outstanding performances. Experiments indicated that our proposed method achieved an accuracy of 87.92%, which is 26.1% higher than that of the traditional CNN method and 1.1% higher than that of the deep convolutional neural network ResNet50.

• International Journal of Computer Science & Network Security
• /
• v.21 no.11
• /
• pp.294-300
• /
• 2021

The current society relies upon social media on an everyday basis, which contributes to finding which of the following supervised machine learning algorithms used in sentiment analysis have higher accuracy in detecting Malay internet slang and short forms which can be offensive to a person. This paper is to determine which of the algorithms chosen in supervised machine learning with higher accuracy in detecting internet slang and short forms. To analyze the results of the supervised machine learning classifiers, we have chosen two types of datasets, one is political topic-based, and another same set but is mixed with 50 tweets per targeted keyword. The datasets are then manually labelled positive and negative, before separating the 275 tweets into training and testing sets. Naïve Bayes and Random Forest classifiers are then analyzed and evaluated from their performances. Our experiment results show that Random Forest is a better classifier compared to Naïve Bayes.

• Journal of Internet Computing and Services
• /
• v.23 no.2
• /
• pp.45-52
• /
• 2022

With the development of artificial intelligence analysis methods, especially machine learning, various fields are widely expanding their application ranges. However, in the case of classical music, there still remain some difficulties in applying machine learning techniques. Genre classification or music recommendation systems generated by deep learning algorithms are actively used in general music, but not in classical music. In this paper, we attempted to classify opera among classical music. To this end, an experiment was conducted to determine which criteria are most suitable among, composer, period of composition, and emotional atmosphere, which are the basic features of music. To generate emotional labels, we adopted zero-shot classification with four basic emotions, 'happiness', 'sadness', 'anger', and 'fear.' After embedding the opera libretto with the doc2vec processing model, the optimal number of clusters is computed based on the result of the elbow method. Decided four centroids are then adopted in k-means clustering to classify unsupervised libretto datasets. We were able to get optimized clustering based on the result of adjusted rand index scores. With these results, we compared them with notated variables of music. As a result, it was confirmed that the four clusterings calculated by machine after training were most similar to the grouping result by period. Additionally, we were able to verify that the emotional similarity between composer and period did not appear significantly. At the end of the study, by knowing the period is the right criteria, we hope that it makes easier for music listeners to find music that suits their tastes.

• Journal of Drive and Control
• /
• v.18 no.4
• /
• pp.9-18
• /
• 2021

This study proposes a method for creating learning datasets to recognize obstacles using deep learning algorithms in automated construction machinery or an autonomous vehicle. Recently, many researchers and engineers have developed various recognition algorithms based on deep learning following an increase in computing power. In particular, the image classification technology and image segmentation technology represent deep learning recognition algorithms. They are used to identify obstacles that interfere with the driving situation of an autonomous vehicle. Therefore, various organizations and companies have started distributing open datasets, but there is a remote possibility that they will perfectly match the user's desired environment. In this study, we created an interface of the virtual simulator such that users can easily create their desired training dataset. In addition, the customized dataset was further advanced by using the RDBMS system, and the recognition rate was improved.

• Journal of the Korea Institute of Military Science and Technology
• /
• v.25 no.5
• /
• pp.443-461
• /
• 2022

Target Detection in synthetic aperture radar(SAR) image is critical for military and national defense. In this paper, we propose YOLOv4-Attention architecture which adds attention modules to YOLOv4 backbone architecture to complement the feature extraction ability for SAR target detection with high accuracy. For training and testing our framework, we present new SAR embedding datasets based on MSTAR SAR public datasets which are about poor environments for target detection such as various clutter, crowded objects, various object size, close to buildings, and weakness of signal-to-clutter ratio. Experiments show that our Attention YOLOv4 architecture outperforms original YOLOv4 architecture in SAR image target detection tasks in poor environments for target detection.

• Journal of IKEEE
• /
• v.27 no.3
• /
• pp.280-287
• /
• 2023

Deep learning advancements can solve computer vision problems, but large-scale datasets are necessary for high accuracy. In this paper, we propose an image generation technique using object bounding boxes and image edge components. The object bounding boxes are extracted from the images through object detection, and image edge components are used as input values for the image generation model to create new image data. As results of experiments, the images generated by the proposed method demonstrated similar image quality to the source images in the image quality assessment, and also exhibited good performance during the deep learning training process.

• Journal of IKEEE
• /
• v.28 no.2
• /
• pp.209-220
• /
• 2024

A sketch is an intuitive means to express information, but compared to actual images, it has the problem of being highly abstract, diverse, and sparse. Recent advances in deep learning models have made it possible to discover features that are common to images and sketches. In this paper, we summarize recent trends in sketch-based image retrieval (SBIR) but it is not limited to SBIR. Besides SBIR, we also introduce sketch-based image recognition and generation studies. Zero-shot learning enables models to recognize categories not encountered during training. Zero-shot SBIR methods are also discussed. Commonly used free-hand sketch datasets are summarized and retrieval performance based on these datasets is reported.

• Journal of Computing Science and Engineering
• /
• v.8 no.4
• /
• pp.207-214
• /
• 2014

In pedestrian detection applications, one of the most popular frameworks that has received extensive attention in recent years is widely known as a 'Hough forest' (HF). To improve the accuracy of detection, this paper proposes a novel split function to exploit the statistical information of the training set stored in each node during the construction of the forest. The proposed split function makes the trees in the forest more robust to noise and illumination changes. Moreover, the errors of each stage in the training forest are minimized using a global loss function to support trees to track harder training samples. After having the forest trained, the standard HF detector follows up to search for and localize instances in the image. Experimental results showed that the detection performance of the proposed framework was improved significantly with respect to the standard HF and alternating decision forest (ADF) in some public datasets.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.12 no.1
• /
• pp.368-391
• /
• 2018

Face recognition (FR) with a single sample per person (SSPP) is common in real-world face recognition applications. In this scenario, it is hard to predict intra-class variations of query samples by gallery samples due to the lack of sufficient training samples. Inspired by the fact that similar faces have similar intra-class variations, we propose a virtual sample generating algorithm called k nearest neighbors based virtual sample generating (kNNVSG) to enrich intra-class variation information for training samples. Furthermore, in order to use the intra-class variation information of the virtual samples generated by kNNVSG algorithm, we propose image set based multimanifold discriminant learning (ISMMDL) algorithm. For ISMMDL algorithm, it learns a projection matrix for each manifold modeled by the local patches of the images of each class, which aims to minimize the margins of intra-manifold and maximize the margins of inter-manifold simultaneously in low-dimensional feature space. Finally, by comprehensively using kNNVSG and ISMMDL algorithms, we propose k nearest neighbor virtual image set based multimanifold discriminant learning (kNNMMDL) approach for single sample face recognition (SSFR) tasks. Experimental results on AR, Multi-PIE and LFW face datasets demonstrate that our approach has promising abilities for SSFR with expression, illumination and disguise variations.

• Smart Structures and Systems
• /
• v.25 no.6
• /
• pp.765-778
• /
• 2020

The present study intends to find a proper solution for the estimation of the physical behaviors of enlarged piles through a combination of small-scale laboratory tests and a hybrid computational predictive intelligence process. In the first step, experimental program is completed considering various critical influential factors. The results of the best multilayer perceptron (MLP)-based predictive network was implemented through three mathematical-based solutions of dragonfly algorithm (DA), whale optimization algorithm (WOA), and ant lion optimization (ALO). Three proposed models, after convergence analysis, suggested excellent performance. These analyses varied based on neurons number (e.g., in the basis MLP hidden layer) and of course, the level of its complexity. The training R² results of the best hybrid structure of DA-MLP, WOA-MLP, and ALO-MLP were 0.996, 0.996, and 0.998 where the testing R² was 0.995, 0.985, and 0.998, respectively. Similarly, the training RMSE of 0.046, 0.051, and 0.034 were obtained for the training and testing datasets of DA-MLP, WOA-MLP, and ALO-MLP techniques, while the testing RMSE of 0.088, 0.053, and 0.053, respectively. This obtained result demonstrates the excellent prediction from the optimized structure of the proposed models if only population sensitivity analysis performs. Indeed, the ALO-MLP was slightly better than WOA-MLP and DA-MLP methods.