• 한국정보디스플레이학회:학술대회논문집
• /
• 2009.10a
• /
• pp.1154-1157
• /
• 2009

This template provides you with an example of the The CIE Color Appearance Model (CIECAM02) is now widely used for various digital imaging systems including digital displays. The CIECAM02 were intended to be an empirical model, however, some aspects of the model are closely related to the human color vision mechanism. This paper will discuss the relationship between human color vision and color imaging.

• Journal of the Korean Society of Manufacturing Technology Engineers
• /
• v.20 no.4
• /
• pp.386-394
• /
• 2011

This paper is concerned with the application of the vision control scheme for robot's point placement task in discontinuous trajectory caused by obstacle. The proposed vision control scheme consists of four models, which are the robot's kinematic model, vision system model, 6-parameters estimation model, and robot's joint angles estimation model. For this study, the discontinuous trajectory by obstacle is divided into two obstacle regions. Each obstacle region consists of 3 cases, according to the variation of number of cameras that can not acquire the vision data. Then, the effects of number of cameras on the proposed robot's vision control scheme are investigated in each obstacle region. Finally, the practicality of the proposed robot's vision control scheme is demonstrated experimentally by performing the robot's point placement task in discontinuous trajectory by obstacle.

• IEMEK Journal of Embedded Systems and Applications
• /
• v.18 no.3
• /
• pp.89-100
• /
• 2023

Recently, on-device artificial intelligence (AI) solutions using mobile devices and embedded edge devices have emerged in various fields, such as computer vision, to address network traffic burdens, low-energy operations, and security problems. Although vision transformer deep learning models have outperformed conventional convolutional neural network (CNN) models in computer vision, they require more computations and parameters than CNN models. Thus, they are not directly applicable to embedded edge devices with limited hardware resources. Many researchers have proposed various model compression methods or lightweight architectures for vision transformers; however, there are only a few studies evaluating the effects of model compression techniques of vision transformers on performance. Regarding this problem, this paper presents a performance evaluation of vision transformers on embedded platforms. We investigated the behaviors of three vision transformers: DeiT, LeViT, and MobileViT. Each model performance was evaluated by accuracy and inference time on edge devices using the ImageNet dataset. We assessed the effects of the quantization method applied to the models on latency enhancement and accuracy degradation by profiling the proportion of response time occupied by major operations. In addition, we evaluated the performance of each model on GPU and EdgeTPU-based edge devices. In our experimental results, LeViT showed the best performance in CPU-based edge devices, and DeiT-small showed the highest performance improvement in GPU-based edge devices. In addition, only MobileViT models showed performance improvement on EdgeTPU. Summarizing the analysis results through profiling, the degree of performance improvement of each vision transformer model was highly dependent on the proportion of parts that could be optimized in the target edge device. In summary, to apply vision transformers to on-device AI solutions, either proper operation composition and optimizations specific to target edge devices must be considered.

• Journal of Korean Academy of Nursing
• /
• v.23 no.4
• /
• pp.678-693
• /
• 1993

The purposes of this study were to describe the level of vision and hearing impairments, depression and functional capacity, among Korean institutionalized elderly and to examine the relation-ship between sensory impairments, depression, and functional capacity in these people. The final pupose was to test the cognitive function path model using sensory competencies as predictors. A convenience sample of thirty nine male and 90 female subjects with a mean age of 80.5 were the subjects of this study. The subjects were tested for cognitive function, and vision and hearing impairments. Physical function and social function were measured by observation of designated task performance by the subjects. Their level of de-pression was measured using a Geriatric Depression Scale administered through an interview. Individual subjective ratings of hearing and vision were marked by the subjects, on a ladder scale. The results of the study showed that 48.8% of the subjects had a hearing impairment, 63.5% had a vision impairement, and 36.4% had both a vision and hearing impairement. The four sensory groups (no sensory impairement, hearing impairement, vision impairement, hearing and vision impairement) were tested for differences in depression, physical function, social behavior and cognitive function. The only significant difference that was found was in cognitive function, between the no sensory impairement group and the hearing and vision impairement group(F=3.25, P<.05), Subjective ratings of hearing showed a significant correlation with cognitive function(r=.34, p<.001) and with social behavior(r=.31, p<.001). There was no correlation between subjective vision ratings and cognitive function or social behavior. However there was a significant correlation between vision and hearing(r=.49, p<.001). There was also a significant negative correlation between age and vision(r=-.21, p<.01) and between age and hear-ing(r=-.34, p<.001). There was a significant correlation between depression and physical function (r=-.32, p<.001) but there was no correlation between depression and cognitive function or social behavior. Based on the literature review and the result, this study, a path model of sensory competence-> cognitive function- >social behavior was developed and tested : Perceived vision and perceived hearing were the exogenous variahles and cognitive function and social behavior were the endogeneous variables in the model. The path analysis result demonstrated an accept-able fit (GFI=.997, AGFI=.972, X$^2$=.72 (p=.396), RMSR=.019) between the data and the model. There was a significant direct effect($\beta$=.38) of perceived hearing on cognitive function. There was a significant direct effect ($\beta$=.32) of cognitive function on social behavior. The total effect of hearing on social behavior was $\beta$=.32 including the indirect effect ($\beta$=.12) . However perceived vsion had little effect ($\beta$=-.08) on cognitive function. The result of path analysis confirms that hearing levels influence cognitive function, and both hearing and cognitive function levels influence social behavior. However, vision has little effect on cognitive function or on social behavior. For the next study, a combined model of the pre viously developed environment - >depression- > physical and social function model, and the present cognitive function model, should be tested to further refine the functional capacity model. There also a need for longitudinal study of functional capacity and sencory competence in order to better understand how declining sensory competence influences functional capacity and how it effects in-creasing dependency and nursing needs in the elderly.

• Transactions of the Korean Society of Machine Tool Engineers
• /
• v.13 no.2
• /
• pp.94-103
• /
• 2004

The vision system model used for this study involves the six parameters that permits a kind of adaptability in that relationship between the camera space location of manipulable visual cues and the vector of robot joint coordinates is estimated in real time. Also this vision control method requires the number of cameras to transform 2-D camera plane from 3-D physical space, and be used irrespective of location of cameras, if visual cues are displayed in the same camera plane. Thus, this study is to investigate the optimal number of cameras used for the developed vision control system according to the change of the number of cameras. This study is processed in the two ways : a) effectiveness of vision system model b) optimal number of cameras. These results show the evidence of the adaptability of the developed vision control method using the optimal number of cameras.

• Journal of the Korean Society of Manufacturing Technology Engineers
• /
• v.26 no.1
• /
• pp.50-58
• /
• 2017

This paper proposes a real-time robotic vision control scheme using the weighting matrix to efficiently process the vision data obtained during robotic movement to a target. This scheme is based on the vision system model that can actively control the camera parameter and robotic position change over previous studies. The vision control algorithm involves parameter estimation, joint angle estimation, and weighting matrix models. To demonstrate the effectiveness of the proposed control scheme, this study is divided into two parts: not applying the weighting matrix and applying the weighting matrix to the vision data obtained while the camera is moving towards the target. Finally, the position accuracy of the two cases is compared by performing the slender bar placement task experimentally.

• The Transactions of The Korean Institute of Electrical Engineers
• /
• v.64 no.7
• /
• pp.1040-1046
• /
• 2015

This paper introduces GP(Genetic Programming) based color detection model for an object detection of humanoid robot vision. Existing color detection methods have used linear/nonlinear transformation of RGB color-model. However, most of cases have difficulties to classify colors satisfactory because of interference of among color channels and susceptibility for illumination variation. Especially, they are outstanding in degraded images from robot vision. To solve these problems, we propose illumination robust and non-parametric multi-colors detection model using evolution of GP. The proposed method is compared to the existing color-models for various environments in robot vision for real humanoid Nao.

• International Journal of Internet, Broadcasting and Communication
• /
• v.13 no.2
• /
• pp.7-13
• /
• 2021

Machine vision is a technology that helps the computer as if a person recognizes and determines things. In recent years, as advanced technologies such as optical systems, artificial intelligence and big data advanced in conventional machine vision system became more accurate quality inspection and it increases the manufacturing efficiency. In machine vision systems using deep learning, the image quality of the input image is very important. However, most images obtained in the industrial field for quality inspection typically contain noise. This noise is a major factor in the performance of the machine vision system. Therefore, in order to improve the performance of the machine vision system, it is necessary to eliminate the noise of the image. There are lots of research being done to remove noise from the image. In this paper, we propose an autoencoder based machine vision system to eliminate noise in the image. Through experiment proposed model showed better performance compared to the basic autoencoder model in denoising and image reconstruction capability for MNIST and fashion MNIST data sets.

• Journal of the Korea Society of Computer and Information
• /
• v.28 no.9
• /
• pp.17-25
• /
• 2023

In this paper, we propose a model that can perform human pose estimation through a MobileViT-based model with fewer parameters and faster estimation. The based model demonstrates lightweight performance through a structure that combines features of convolutional neural networks with features of Vision Transformer. Transformer, which is a major mechanism in this study, has become more influential as its based models perform better than convolutional neural network-based models in the field of computer vision. Similarly, in the field of human pose estimation, Vision Transformer-based ViTPose maintains the best performance in all human pose estimation benchmarks such as COCO, OCHuman, and MPII. However, because Vision Transformer has a heavy model structure with a large number of parameters and requires a relatively large amount of computation, it costs users a lot to train the model. Accordingly, the based model overcame the insufficient Inductive Bias calculation problem, which requires a large amount of computation by Vision Transformer, with Local Representation through a convolutional neural network structure. Finally, the proposed model obtained a mean average precision of 0.694 on the MS COCO benchmark with 3.28 GFLOPs and 9.72 million parameters, which are 1/5 and 1/9 the number compared to ViTPose, respectively.

• Korean Journal of Radiology
• /
• v.25 no.4
• /
• pp.403-406
• /
• 2024