• The Journal of the Convergence on Culture Technology
• /
• v.10 no.3
• /
• pp.909-914
• /
• 2024

This study suggests a quantitative evaluation of transfer learning, which is widely used in various AI fields, including image recognition for robot vision. Quantitative and qualitative analyses of results applying transfer learning are presented, but transfer learning itself is not discussed. Therefore, this study proposes a quantitative evaluation of transfer learning itself based on MNIST, a handwritten digit database. For the reference network, the change in recognition accuracy according to the depth of the transfer learning frozen layer and the ratio of transfer learning data and pre-training data is tracked. It is observed that when freezing up to the first layer and the ratio of transfer learning data is more than 3%, the recognition accuracy of more than 90% can be stably maintained. The transfer learning quantitative evaluation method of this study can be used to implement transfer learning optimized according to the network structure and type of data in the future, and will expand the scope of the use of robot vision and image analysis AI in various environments.

• Smart Media Journal
• /
• v.5 no.4
• /
• pp.63-68
• /
• 2016

Texture transfer is a method to transfer the texture of an input image into a target image, and is also used for transferring artistic style of the input image. This study presents a real-time texture transfer for generating artistic style video. In order to enhance performance, this paper proposes a parallel framework using T-shape kernel used in general texture transfer on GPU. To accelerate motion computation time which is necessarily required for maintaining temporal coherence, a multi-scaled motion field is proposed in parallel concept. Through these approach, an artistic texture transfer for video with a real-time performance is archived.

• Journal of IKEEE
• /
• v.26 no.4
• /
• pp.639-646
• /
• 2022

In this paper, to solve the large dataset problem, we collect images through an image collection method called web crawling and build datasets for use in image classification models through a data preprocessing process. We also propose a lightweight model that can automatically classify images by adding category values by incorporating transfer learning into the image classification model and an image classification model that reduces training time and achieves high accuracy.

• Journal of Information Processing Systems
• /
• v.19 no.5
• /
• pp.663-672
• /
• 2023

Most vehicle detection methods have poor vehicle feature extraction performance at night, and their robustness is reduced; hence, this study proposes a night vehicle detection method based on style transfer image enhancement. First, a style transfer model is constructed using cycle generative adversarial networks (cycleGANs). The daytime data in the BDD100K dataset were converted into nighttime data to form a style dataset. The dataset was then divided using its labels. Finally, based on a YOLOv5s network, a nighttime vehicle image is detected for the reliable recognition of vehicle information in a complex environment. The experimental results of the proposed method based on the BDD100K dataset show that the transferred night vehicle images are clear and meet the requirements. The precision, recall, mAP@.5, and mAP@.5:.95 reached 0.696, 0.292, 0.761, and 0.454, respectively.

• Knowledge Management Research
• /
• v.12 no.3
• /
• pp.27-37
• /
• 2011

It is not uncommon to witness brand image transference which is a development of corporate image from a collection of individual brand images. The brand image transfer process is the influence of consumer attitudes toward certain brands on overall evaluation of the company. To understand the image transfer process, we examine the influence of brand level evaluation on overall company evaluation through food and beverage consumer products in China, where active competitions among global brands exist.

• Proceedings of the KSME Conference
• /
• 2001.11b
• /
• pp.711-716
• /
• 2001

This paper presents an information about the heat transfer characteristics of impinging jet in eletronic equipment with infrared image processing unit. There have been many experimental investigations and theoretical studies on impinging jet because of application in a wide variety of industrial process including electronic equipment. In this study, we used infrared image processing unit to visualize heat transfer characteristics of impinging jet in electronic equipment. Infrared image processing unit is one of non-contact temperature measuring methods and it is possible to minimize flow resistance and this measurement is comparatively accurate. The main parameters are nozzle exit angle $(30^{\circ},\;45^{\circ},\;60^{\circ},\;75^{\circ},\;90^{\circ})$ and distance between nozzle and heat source is fixed 6d("d" is diameter of circular nozzle(10 mm). Reynolds number is 4500.

• Journal of Astronomy and Space Sciences
• /
• v.30 no.4
• /
• pp.321-326
• /
• 2013

The modulation transfer function (MTF) is a widely used indicator in assessments of remote-sensing image quality. This MTF method is also used to restore information to a standard value to compensate for image degradation caused by atmospheric or satellite jitter effects. In this study, we evaluated MTF values as an image quality indicator for the Geostationary Ocean Color Imager (GOCI). GOCI was launched in 2010 to monitor the ocean and coastal areas of the Korean peninsula. We evaluated in-orbit MTF value based on the GOCI image having a 500-m spatial resolution in the first time. The pulse method was selected to estimate a point spread function (PSF) with an optimal natural target such as a Seamangeum Seawall. Finally, image restoration was performed with a Wiener filter (WF) to calculate the PSF value required for the optimal regularization parameter. After application of the WF to the target image, MTF value is improved 35.06%, and the compensated image shows more sharpness comparing with the original image.

• Journal of Broadcast Engineering
• /
• v.27 no.1
• /
• pp.104-123
• /
• 2022

Style transfer based on neural network provides very high quality results by reflecting the high level structural characteristics of images, and thereby has recently attracted great attention. This paper deals with the problem of resolution limitation due to GPU memory in performing such neural style transfer. We can expect that the gradient operation for style transfer based on partial image, with the aid of the fixed size of receptive field, can produce the same result as the gradient operation using the entire image. Based on this idea, each component of the style transfer loss function is analyzed in this paper to obtain the necessary conditions for partitioning and padding, and to identify, among the information required for gradient calculation, the one that depends on the entire input. By structuring such information for using it as auxiliary constant input for partition-based gradient calculation, this paper develops a recursive algorithm for super high-resolution image style transfer. Since the proposed method performs style transfer by partitioning input image into the size that a GPU can handle, it can perform style transfer without the limit of the input image resolution accompanied by the GPU memory size. With the aid of such super high-resolution support, the proposed method can provide a unique style characteristics of detailed area which can only be appreciated in super high-resolution style transfer.

• Journal of Korean Dental Science
• /
• v.4 no.1
• /
• pp.20-25
• /
• 2011

Purpose: To describe the newly developed Film image transfer system (FITS) for proper positioning of the orthodontic mini-implant in the narrow interdental space and considerations for better application. Materials and Methods: A patient who was planning to have orthodontic mini-implant treatment on the posterior maxilla was recruited to assess the feasibility of FITS. Dental radiographic film and bite record was taken. And then the film image was transferred on the photographic emulsion coated model using transfer light through film projector (enlarger). After exposing the photo emulsion coating on the model, the image was developed with a working solution for a paper developer and fixed. The surgical guide for the mini-implant was fabricated from the transported FITS data. Results: The completed surgical guide was easily placed intraorally, and allowed a simple and rapid placement of the mini-implant. The site of the implant placement was accurate as planned position. Conclusion: In the reported case, The FITS technique represents an effort to minimize risk to the patient and produce consistently good results based upon accurate information about the anatomy of the implant site.

• Journal of Biomedical Engineering Research
• /
• v.17 no.1
• /
• pp.33-42
• /
• 1996

A multimedia remote ultrasound image transfer system for tole-medical services was implemented. The image data of the ultrasound scanner system of the slave is compressed using image compression system. It is transferred to the master PC via ISDN UNI(User-Network Interface) and decompressd by image decompression system. The total system is composed of three parts; first, the image capture card which transfers bidirection- al data between ultrasound scanner system and PC, second, image compression and decompression card, finally, ISDN TA(Terminal Adaptor) card for transferring the image. This system has a easy user interface because it is executed on the basis of MS-WINDOWS. So it is capable of serving medical services at a remote place.