• The Journal of The Korea Institute of Intelligent Transport Systems
• /
• v.21 no.5
• /
• pp.274-286
• /
• 2022

Multi-object tracking has been studied for a long time under computer vision and plays a critical role in applications such as autonomous driving and driving assistance. Multi-object tracking techniques generally consist of a detector that detects objects and a tracker that tracks the detected objects. Various publicly available datasets allow us to train a detector model without much effort. However, there are relatively few publicly available datasets for training a tracker model, and configuring own tracker datasets takes a long time compared to configuring detector datasets. Hence, the detector is often developed separately with a tracker module. However, the separated tracker should be adjusted whenever the former detector model is changed. This study proposes a system that can train a model that performs detection and tracking simultaneously using only the detector training datasets. In particular, a Siam network with augmentation is used to compose the detector and tracker. Experiments are conducted on public datasets to verify that the proposed algorithm can formulate a real-time multi-object tracker comparable to the state-of-the-art tracker models.

• Journal of the Korea institute for structural maintenance and inspection
• /
• v.28 no.4
• /
• pp.55-61
• /
• 2024

Recently, the number of aging concrete structures is steadily increasing. This is because many of these structures are reaching their expected lifespan. Such structures require accurate inspections and persistent maintenance. Otherwise, their original functions and performance may degrade, potentially leading to safety accidents. Therefore, research on objective inspection technologies using deep learning and computer vision is actively being conducted. High-resolution images can accurately observe not only micro cracks but also spalling and exposed rebar, and deep learning enables automated detection. High detection performance in deep learning is only guaranteed with diverse and numerous training datasets. However, surface damage to concrete is not commonly captured in images, resulting in a lack of training data. To overcome this limitation, this study proposed a method for generating concrete surface damage images, including cracks, spalling, and exposed rebar, using stable diffusion. This method synthesizes new damage images by paired text and image data. For this purpose, a training dataset of 678 images was secured, and fine-tuning was performed through low-rank adaptation. The quality of the generated images was compared according to three base models of stable diffusion. As a result, a method to synthesize the most diverse and high-quality concrete damage images was developed. This research is expected to address the issue of data scarcity and contribute to improving the accuracy of deep learning-based damage detection algorithms in the future.

• Convergence Security Journal
• /
• v.22 no.3
• /
• pp.25-32
• /
• 2022

Recently, the image processing industry has been activated as deep learning-based technology is introduced in the image recognition and detection field. With the development of deep learning technology, learning model vulnerabilities for adversarial attacks continue to be reported. However, studies on countermeasures against poisoning attacks that inject malicious data during learning are insufficient. The conventional countermeasure against poisoning attacks has a limitation in that it is necessary to perform a separate detection and removal operation by examining the training data each time. Therefore, in this paper, we propose a technique for reducing the attack success rate by applying modifications to the training data and inference data without a separate detection and removal process for the poison data. The One-shot kill poison attack, a clean label poison attack proposed in previous studies, was used as an attack model. The attack performance was confirmed by dividing it into a general attacker and an intelligent attacker according to the attacker's attack strategy. According to the experimental results, when the proposed defense mechanism is applied, the attack success rate can be reduced by up to 65% compared to the conventional method.

• Journal of radiological science and technology
• /
• v.26 no.3
• /
• pp.33-39
• /
• 2003

Purpose : Be aware of clinical possibilities on image quality by comparison of contrast-enhanced dynamic CT and MR imaging applied of MIP technique after the experimentally induced clonorchasis infection in dogs. Materials and Method : Twenty mongrel dogs prepared in zoo-laboratory were followed up with serial CT scans and MR imaging for 13 weeks after the experimental infection in liver. Two-phase helical CT was acquired in the supine position with the following scanning parameters. After the injection of contrast material, the arterial phase was initiated using a bolus-racking method. The portal phase scan was started 15 seconds after the arterial phase scan. CT protocol was determined after single level dynamic scans. MR imaging used the CP body coil and images get a 2D image using HASTE, FLASH, TSE pulse sequence. Bile duct MR imaging were obtained in three plans. Then each image was post processed by using target MIP algorithm. Two experimentation above, as a method of evaluation, one pathologist, three radiologist and five radiological technologist were analyzed visually for evaluation of following findings, enhancement of the bile duct wall, dilatation of bile duct tip, liver parenchyma, background suppression. Results : Five dogs was died of a disease after the infection, the rest one else shows the chronic dilatation of the intrahepatic bile duct with CT and MR imaging. Contrast administration of CT shows the contrast-enhanced of the bile duct walls with live parenchyma. MR imaging calculated of CNR and CR from pulse sequence for comparative evaluation and shows the pattern of the intrahepatic bile duct, dilatation of bile duct tip using MIP technique. CNR of the clonorchiasis, HASTE was $16{\pm}0.83$, TSE $7.06{\pm}3.0$, FLASH $1.19{\pm}0.2$ and CR, HASTE was 73.3%, TSE 62.3%, FLASH 6.4%. Conclusion : CT and MR imaging is very usefulness in diagnosis of dog clonorchiasis.

• Investigative Magnetic Resonance Imaging
• /
• v.18 no.1
• /
• pp.34-42
• /
• 2014

Purpose : To evaluate the correlation of lesion-to-normal ratio (LNR) of signal intensity from double inversion recovery MR imaging and total choline-containing compound (tCho) resonance from single voxel MR spectroscopy in breast cancers. Materials and Methods: Between August 2008 and December 2009, 28 patients who were diagnosed as breast cancer and had undergone both double inversion recovery (DIR) MR imaging and MR spectroscopy (MRS) were included in this study. The signal intensities of the lesion (L) and ipsilateral normal breast tissue (N) were measured in region of interest of each breast cancer in DIR and contrast enhance MR image (CE-T1WI) to calculate the LNR value for each technique. MRS was performed using single-voxel MR spectroscopy. The height, width and area of tCho resonance were compared with each LNR of DIR and CE-T1WI. We used Pearson's correlation coefficient(r) for correlation analysis and the significance level was p=0.05. Results: There was no statistically significant correlation between LNR of CE-T1WI and height (r=-0.322, p=0.094), width (r=-0.233, p=0.232) and area (r=-0.309, p=0.109) of MRS tCho. There was no statistically significant correlation between LNR of DIR and height (r=0.067, p=0.735), width (r=-0.287, p=0.139) and area (r=0.012, p=0.953) of MRS tCho, either. The Pearson's correlation coefficient was 0.186 between LNRs of CET1WI and DIR (p=0.344). Conclusion: There was no statistically significant correlation between LNR of DIR and relative amount of tCho resonance of MRS.

• Archives of design research
• /
• v.18 no.3 s.61
• /
• pp.85-94
• /
• 2005

Most of product designers use 3D CAD system as a inevitable design tool nowadays and many new products are developed through a concurrent engineering process. However, it is very difficult for novice designers to get the sense of reality from modeling objects shown in the computer screens. Such a intangibility problem comes from the lack of haptic interactions and contextual information about the real space because designers tend to do 3D modeling works only in a virtual space of 3D CAD system. To address this problem, this research investigate the possibility of a interactive quantified structure simulation for product design using AR(augmented reality) which can register a 3D CAD modeling object on the real space. We built a quantified structure simulation system based on AR and conducted a series of experiments to measure how accurately human perceive and adjust the size of virtual objects under varied experimental conditions in the AR environment. The experiment participants adjusted a virtual cube to a reference real cube within 1.3% relative error(5.3% relative StDev). The results gave the strong evidence that the participants can perceive the size of a virtual object very accurately. Furthermore, we found that it is easier to perceive the size of a virtual object in the condition of presenting plenty of real reference objects than few reference objects, and using LCD panel than HMD. We tried to apply the simulation system to identify preference characteristics for the appearance design of a home-service robot as a case study which explores the potential application of the system. There were significant variances in participants' preferred characteristics about robot appearance and that was supposed to come from the lack of typicality of robot image. Then, several characteristic groups were segmented by duster analysis. On the other hand, it was interesting finding that participants have significantly different preference characteristics between robot with arm and armless robot and there was a very strong correlation between the height of robot and arm length as a human body.

• KIPS Transactions on Computer and Communication Systems
• /
• v.5 no.10
• /
• pp.293-302
• /
• 2016

Since the electricity is produced and consumed simultaneously, predicting the electric load and securing affordable electric power are necessary for reliable electric power supply. In particular, a university campus is one of the highest power consuming institutions and tends to have a wide variation of electric load depending on time and environment. For these reasons, an accurate electric load forecasting method that can predict power consumption in real-time is required for efficient power supply and management. Even though various influencing factors of power consumption have been discovered for the educational institutions by analyzing power consumption patterns and usage cases, further studies are required for the quantitative prediction of electric load. In this paper, we build an electric load forecasting model by implementing and evaluating various machine learning algorithms. To do that, we consider three building clusters in a campus and collect their power consumption every 15 minutes for more than one year. In the preprocessing, features are represented by considering periodic characteristic of the data and principal component analysis is performed for the features. In order to train the electric load forecasting model, we employ both artificial neural network and support vector machine. We evaluate the prediction performance of each forecasting model by 5-fold cross-validation and compare the prediction result to real electric load.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.22 no.2
• /
• pp.63-70
• /
• 2022

Although 2D Object detection has been largely improved in the past years with the advance of deep learning methods and the use of large labeled image datasets, 3D object detection from 2D imagery is a challenging problem in a variety of applications such as robotics, due to the lack of data and diversity of appearances and shapes of objects within a category. Google has just announced the launch of Objectron that has a novel data pipeline using mobile augmented reality session data. However, it also is corresponding to 2D-driven 3D object detection technique. This study explores more mature 2D object detection method, and applies its 2D projection to Objectron 3D lifting system. Most object detection methods use bounding boxes to encode and represent the object shape and location. In this work, we explore a stochastic representation of object regions using Gaussian distributions. We also present a similarity measure for the Gaussian distributions based on the Hellinger Distance, which can be viewed as a stochastic Intersection-over-Union. Our experimental results show that the proposed Gaussian representations are closer to annotated segmentation masks in available datasets. Thus, less accuracy problem that is one of several limitations of Objectron can be relaxed.

• Trans-
• /
• v.15
• /
• pp.1-24
• /
• 2023

This study examines the application of data visualization technology using a clustered data algorithm called 'Tamed Cloud' to virtual spaces and seeks the possibility of implementing it in various types of realistic exhibition contents. To this end, we first attempt to classify virtual reality (VR) exhibition contents starting with COVID-19, and summarize technologies applied. Also, various realistic exhibition contents provide visitors with an opportunity to appreciate the artworks through online and virtual exhibitions. In this trend, virtual reality and augmented reality (AR) technologies have been introduced, allowing visitors to enjoy the artwork more immersively, and the possibility of realistic exhibition content with interaction between the artwork and the user is also being demonstrated. Based on this background, this study examines the history of exhibition contents by dividing them before and after the advent of virtual reality technology, and examines how the clustered algorithm technology called Tamed Cloud was applied to virtual space and implemented as a realistic exhibition content in <70mK> project. By synthesizing all of this, we propose a convergence method of data visualization, virtual reality, and realistic content, and propose it as a new alternative to realistic exhibition content in virtual space.

• Korean Journal of Remote Sensing
• /
• v.39 no.2
• /
• pp.143-156
• /
• 2023

To precisely and reliably analyze the contents of the satellite imagery, recognizing the clouds which are the obstacle to gathering the useful information is essential. In recent times, deep learning yielded satisfactory results in various tasks, so many studies using deep neural networks have been conducted to improve the performance of cloud detection. However, existing methods for cloud detection have the limitation on increasing the performance due to the adopting the network models for semantic image segmentation without modification. To tackle this problem, we introduced the multi-branch neural architecture search to find optimal network structure for cloud detection. Additionally, the proposed method adopts the soft intersection over union (IoU) as loss function to mitigate the disagreement between the loss function and the evaluation metric and uses the various data augmentation methods. The experiments are conducted using the cloud detection dataset acquired by Arirang-3/3A satellite imagery. The experimental results showed that the proposed network which are searched network architecture using cloud dataset is 4% higher than the existing network model which are searched network structure using urban street scenes with regard to the IoU. Also, the experimental results showed that the soft IoU exhibits the best performance on cloud detection among the various loss functions. When comparing the proposed method with the state-of-the-art (SOTA) models in the field of semantic segmentation, the proposed method showed better performance than the SOTA models with regard to the mean IoU and overall accuracy.