• Journal of Information Processing Systems
• /
• v.14 no.1
• /
• pp.205-217
• /
• 2018

Deep learning using convolutional neural networks (CNNs) is being studied in various fields of image recognition and these studies show excellent performance. In this paper, we compare the performance of CNN architectures, KCR-AlexNet and KCR-GoogLeNet. The experimental data used in this paper is obtained from PHD08, a large-scale Korean character database. It has 2,187 samples of each Korean character with 2,350 Korean character classes for a total of 5,139,450 data samples. In the training results, KCR-AlexNet showed an accuracy of over 98% for the top-1 test and KCR-GoogLeNet showed an accuracy of over 99% for the top-1 test after the final training iteration. We made an additional Korean character dataset with fonts that were not in PHD08 to compare the classification success rate with commercial optical character recognition (OCR) programs and ensure the objectivity of the experiment. While the commercial OCR programs showed 66.95% to 83.16% classification success rates, KCR-AlexNet and KCR-GoogLeNet showed average classification success rates of 90.12% and 89.14%, respectively, which are higher than the commercial OCR programs' rates. Considering the time factor, KCR-AlexNet was faster than KCR-GoogLeNet when they were trained using PHD08; otherwise, KCR-GoogLeNet had a faster classification speed.

• The Bulletin of The Korean Astronomical Society
• /
• v.42 no.2
• /
• pp.60.2-61
• /
• 2017

A Convolutional Neural Network(CNN) is one of the well-known deep-learning methods in image processing and computer vision area. In this study, we apply CNN to two kinds of flare forecasting models: flare classification and occurrence. For this, we consider several pre-trained models (e.g., AlexNet, GoogLeNet, and ResNet) and customize them by changing several options such as the number of layers, activation function, and optimizer. Our inputs are the same number of SOHO)/MDI images for each flare class (None, C, M and X) at 00:00 UT from Jan 1996 to Dec 2010 (total 1600 images). Outputs are the results of daily flare forecasting for flare class and occurrence. We build, train, and test the models on TensorFlow, which is well-known machine learning software library developed by Google. Our major results from this study are as follows. First, most of the models have accuracies more than 0.7. Second, ResNet developed by Microsoft has the best accuracies : 0.77 for flare classification and 0.83 for flare occurrence. Third, the accuracies of these models vary greatly with changing parameters. We discuss several possibilities to improve the models.

• IEMEK Journal of Embedded Systems and Applications
• /
• v.17 no.4
• /
• pp.217-228
• /
• 2022

Currently, there are increasing demands for applying deep neural networks (DNNs) in the embedded domain such as classification and object detection. The DNN processing in embedded domain often requires custom hardware such as NPU for acceleration due to the constraints in power, performance, and area. Processing DNN models requires a large amount of data, and its seamless transfer to NPU is crucial for performance. In this paper, we developed a cycle-accurate NPU simulator to evaluate diverse NPU microarchitectures. In addition, we propose a novel technique for reducing the number of memory accesses when processing convolutional layers in convolutional neural networks (CNNs) on the NPU. The main idea is to reuse data with memory interleaving, which recycles the overlapping data between previous and current input windows. Data memory interleaving makes it possible to quickly read consecutive data in unaligned locations. We implemented the proposed technique to the cycle-accurate NPU simulator and measured the performance with LeNet-5, VGGNet-16, and ResNet-50. The experiment shows up to 2.08x speedup in processing one convolutional layer, compared to the baseline.

• Korean Journal of Remote Sensing
• /
• v.37 no.5_1
• /
• pp.1149-1161
• /
• 2021

With a trend of the utilization of computer vision for satellite images, cloud detection using deep learning also attracts attention recently. In this study, we conducted a U-Net cloud detection modeling using SPARCS (Spatial Procedures for Automated Removal of Cloud and Shadow) Cloud Dataset with the image data augmentation and carried out 10-fold cross-validation for an objective assessment of the model. Asthe result of the blind test for 1800 datasets with 512 by 512 pixels, relatively high performance with the accuracy of 0.821, the precision of 0.847, the recall of 0.821, the F1-score of 0.831, and the IoU (Intersection over Union) of 0.723. Although 14.5% of actual cloud shadows were misclassified as land, and 19.7% of actual clouds were misidentified as land, this can be overcome by increasing the quality and quantity of label datasets. Moreover, a state-of-the-art DeepLab V3+ model and the NAS (Neural Architecture Search) optimization technique can help the cloud detection for CAS500 (Compact Advanced Satellite 500) in South Korea.

• Journal of KIISE
• /
• v.44 no.1
• /
• pp.44-50
• /
• 2017

Pedestrian detection, an important component of autonomous driving and driving assistant system, has been extensively studied for many years. In particular, image based pedestrian detection methods such as Hierarchical classifier or HOG and, deep models such as ConvNet are well studied. The evaluation score has increased by the various methods. However, pedestrian detection requires high sensitivity to errors, since small error can lead to life or death problems. Consequently, further reduction in pedestrian detection error rate of autonomous systems is required. We proposed a new method to detect pedestrians and reduce the error rate by using the Faster R-CNN with new developed pedestrian training data sets. Finally, we compared the proposed method with the previous models, in order to show the improvement of our method.

• KSCE Journal of Civil and Environmental Engineering Research
• /
• v.40 no.2
• /
• pp.239-245
• /
• 2020

Recently, BIM (Building Information Modeling) are widely being utilized in Construction industry. However, most structures that have been constructed in the past do not have BIM. For structures without BIM, the use of SfM (Structure from Motion) techniques in the 2D image obtained from the camera allows the generation of 3D model point cloud data and BIM to be established. However, since these generated point cloud data do not contain semantic information, it is necessary to manually classify what elements of the structure. Therefore, in this study, deep learning was applied to automate the process of classifying structural components. In the establishment of deep learning network, Inception-ResNet-v2 of CNN (Convolutional Neural Network) structure was used, and the components of bridge structure were learned through transfer learning. As a result of classifying components using the data collected to verify the developed system, the components of the bridge were classified with an accuracy of 96.13 %.

• International Journal of Computer Science & Network Security
• /
• v.21 no.12spc
• /
• pp.463-468
• /
• 2021

Single image super-resolution (SISR) is an image processing technique, and its main target is to reconstruct the high-quality or high-resolution (HR) image from the low-quality or low-resolution (LR) image. Currently, deep learning-based convolutional neural network (CNN) image super-resolution approaches achieved remarkable improvement over the previous approaches. Furthermore, earlier approaches used hand designed filter to upscale the LR image into HR image. The design architecture of such approaches is easy, but it introduces the extra unwanted pixels in the reconstructed image. To resolve these issues, we propose novel deep learning-based approach known as Lightweight deep CNN-based approach for Single Image Super-Resolution (LDCSIR). In this paper, we propose a new architecture which is inspired by ResNet with Inception blocks, which significantly drop the computational cost of the model and increase the processing time for reconstructing the HR image. Compared with the other state of the art methods, LDCSIR achieves better performance in terms of quantitively (PSNR/SSIM) and qualitatively.

• Journal of IKEEE
• /
• v.27 no.4
• /
• pp.439-449
• /
• 2023

In an era marked by the increasing use of drones and the growing demand for indoor surveillance, the development of a robust application for detecting and tracking both drones and humans within indoor spaces becomes imperative. This study presents an innovative application that uses FMCW radar to detect human and drone motions from the cloud point. At the outset, the DBSCAN (Density-based Spatial Clustering of Applications with Noise) algorithm is utilized to categorize cloud points into distinct groups, each representing the objects present in the tracking area. Notably, this algorithm demonstrates remarkable efficiency, particularly in clustering drone point clouds, achieving an impressive accuracy of up to 92.8%. Subsequently, the clusters are discerned and classified into either humans or drones by employing a deep learning model. A trio of models, including Deep Neural Network (DNN), Residual Network (ResNet), and Long Short-Term Memory (LSTM), are applied, and the outcomes reveal that the ResNet model achieves the highest accuracy. It attains an impressive 98.62% accuracy for identifying drone clusters and a noteworthy 96.75% accuracy for human clusters.

• Journal of Korea Multimedia Society
• /
• v.21 no.8
• /
• pp.960-968
• /
• 2018

The mobile visual search service uses a query image to acquire linkage information through pre-constructed DB search. From the standpoint of this purpose, it would be more useful if you could perform a search on a web-based keyword search system instead of a pre-built DB search. In this paper, we propose a representative query extraction algorithm to be used as a keyword on a web-based search system. To do this, we use image classification labels generated by the CNN (Convolutional Neural Network) algorithm based on Deep Learning, which has a remarkable performance in image recognition. In the query extraction algorithm, dictionary meaningful words are extracted using Wikipedia, and hierarchical categories are constructed using WordNet. The performance of the proposed algorithm is evaluated by measuring the system response time.

• Journal of Internet Computing and Services
• /
• v.18 no.5
• /
• pp.61-67
• /
• 2017

In this paper we introduce two experimental results from classification of Instagram images and some valuable lessons from them. We have tried some experiments for evaluating the competitive power of Convolutional Neural Network(CNN) in classification of real social network images such as Instagram images. We used AlexNet and ResNet, which showed the most outstanding capabilities in ImageNet Large Scale Visual Recognition Challenge(ILSVRC) 2012 and 2015, respectively. And we used 240 Instagram images and 12 pre-defined categories for classifying social network images. Also, we performed fine-tuning using Inception V3 model, and compared those results. In the results of four cases of AlexNet, ResNet, Inception V3 and fine-tuned Inception V3, the Top-1 error rates were 49.58%, 40.42%, 30.42%, and 5.00%. And the Top-5 error rates were 35.42%, 25.00%, 20.83%, and 0.00% respectively.