• 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 KIISE:Software and Applications
• /
• v.28 no.9
• /
• pp.649-661
• /
• 2001

While the conventional edge detection can be considered as the problem of determining the existence of edges at certain locations, the fuzzy edge modeling can be considered as the problem of determining the membership values of edges. Thus, if the location of an edge is unclear, or if the intensity function is different from the ideal edge model, the degree of edgeness at the location is represented as a fuzzy membership value. Using the concept of fuzzy edgeness, an automatic smoothing parameter evaluation and selection method for a conventional edge detector is proposed. This evaluation method uses the fuzzy edge modeling, and can analyze the effect of smoothing parameter to determine an optimal parameter for a given image. By using the selected parameter we can detect least ambiguous edges of a detection method for an image. The effectiveness of the parameter evaluation method is analyzed and demonstrated using a set of synthetic and real images.

• Convergence Security Journal
• /
• v.7 no.3
• /
• pp.39-44
• /
• 2007

In order to detect and locate edge features precisely in real images we have developed an algorithm by introducing a nonlocal differentiation of intensity profiles called adaptive directional derivative (ADD), which is evaluated independently of varying ramp widths. In this paper, we first develop the edge detector system employing the ADD and then, the performance of the algorithm is illustrated by comparing the results to those from the Canny's edge detector.

• Journal of information and communication convergence engineering
• /
• v.12 no.3
• /
• pp.193-197
• /
• 2014

Edges are a robust feature for object detection. In this paper, we present an edge-based background modeling method for the detection of moving objects. The edges in the image frames were mapped using robust Canny edge detector. Two edge maps were created and combined to calculate the ultimate moving-edge map. By selecting all the edge pixels of the current frame above the defined threshold of the ultimate moving edges, a temporary background-edge map was created. If the frequencies of the temporary background edge pixels for several frames were above the threshold, then those edge pixels were treated as background edge pixels. We conducted a performance comparison with previous works. The existing edge-based moving-object detection algorithms pose some difficulty due to the changes in background motion, object shape, illumination variation, and noises. The result of the performance evaluation shows that the proposed algorithm can detect moving objects efficiently in real-world scenarios.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.25 no.6B
• /
• pp.993-1001
• /
• 2000

This research features a method that quantitatively evaluates the performance of edge detection algorithms. Contrary to conventional methods that evaluate the performance of edge detection as a function of the amount of noise added to he input image, the proposed method is capable of assessing the performance of edge detection algorithms based on chosen parameters that influence the performance of edge detection. We have proposed a quantitative measure, called average performance index, that compares the average performance of different edge detection algorithms. We have applied the method to the commonly used edge detectors, Sobel, LOG(Laplacian of Gaussian), and Canny edge detectors for noisy images that contain straight line edges and curved line edges. Two kinds of noises i.e, Gaussian and impulse noises, are used. Experimental results show that our method of quantitatively evaluating the performance of edge detection algorithms can facilitate the selection of the optimal dge detection algorithm for a given task.

• Journal of radiological science and technology
• /
• v.43 no.1
• /
• pp.15-21
• /
• 2020

This study was purpose to evaluation of Modulation Transfer Function in Measurements by using the International electrotechnical commission standard(IEC 62220-1) which were edge device each angle by using edge method. In this study was Aero(Konica, Japan) image receptor which is a indirect Flat panel detector(FPD) was used. The size of matrix 1994 × 2430 (14"× 17" inch) which performed 12 bit processing and pixel pitch is 175 ㎛. The results of shown as MTF measurements at IEC standard. The amount of data seemed reasonable and at an MTF value of 0.1 the spatial frequencies were 2.56 cycles/mm at an angle of 2.4°. MTF value of 0.5 the spatial frequencies were 1.32 cycles/mm at an angle of 2.4°. This study were to evaluate MTF by setting each angle 2.0°~2.8° degrees the most effective optimal edge angle and to suggest the quantitative methods of measuring by using IEC.

• 한국전산유체공학회:학술대회논문집
• /
• 2000.10a
• /
• pp.32-38
• /
• 2000

Wing flap deflection angles of a supersonic transport are optimized to improve transonic cruise performance. For this end, a numerical optimization method is adopted using a three-dimensional unstructured Euler code and a discrete adjoint code. Deflection angles of ten flaps; five for leading edge and five fur railing edge, are employed as design variables. The elliptic equation method is adopted for the interior grid modification during the design process. Interior grid sensitivities are neglected for efficiency. Also tested is the validity of the approximate gradient evaluation method for the present design problem and found that it is applicable for loading edge flap design in cases of no shock waves on the wing surface. The BFGS method is used to minimize the drag with constraints on the lift and upper surface Mach numbers. Two design examples are conducted; one is leading edge flap design, and the other is simultaneous design of leading edge and trailing edge flaps. The latter gave a smaller drag than the former by about two counts.

• Journal of Mechanical Science and Technology
• /
• v.18 no.3
• /
• pp.481-490
• /
• 2004

Steady state flow calculations are executed for turbo-pump inducers of modern design to validate the performance of Tascflow code. Hydrodynamic performance of inducers is evaluated and structure of the passage flow and leading edge recirculation are also investigated. Calculated results show good coincidence with experimental data of static pressure performance and velocity profiles over the leading edge. Upstream recirculation, tip leakage and vortex flow at the blade tip and near leading edge are main sources of pressure loss. Amount of pressure loss from the upstream to the leading edge corresponds to that of whole pressure loss through the blade passage. The viscous loss is considerably large due to the strong secondary flow. There appears more stronger leading edge recirculation for the backswept inducer, and this increases the pressure loss. However, blade loading near the leading edge is considerably reduced and cavitation inception delayed.

• Proceedings of the Korean Society of Computer Information Conference
• /
• 2018.07a
• /
• pp.133-134
• /
• 2018

본 논문에서는 Edge Computing 알고리즘의 성능 분석 및 평가를 위해 SDN 무선 테스트베드를 구현하는 방안을 제안한다. Edge Computing 환경은 수 많은 노드를 연결해 테스트해야 하며, 따라서 무선 테스트베드가 적합하다. Edge Computing 기술이 발전함에 따라 네트워크가 점점 더 복잡해지고 정확하고 확장 가능한 테스트베드 기술에 대한 필요성이 급증하고 있다. 네트워크 연구를 위해 대규모 테스트 베드가 필수적이다. 본 논문에서는 Raspberry-Pi를 이용해 Wireless SDN 테스트베드 구축 방안에 대해 소개하고자 한다.

• 한국방재학회:학술대회논문집
• /
• 2008.02a
• /
• pp.333-336
• /
• 2008

U-Channel Bridge is effective bridge type, because its edge beam performs role of barrier and enables to reduce additional dead loads. Nevertheless, there is possibility of bridge collapse under impact load due to car crash. Also, edge beam must have ability to induce safe driving and prevent falling accidents. Therefore, this study carries out analysis of behavior of edge beam and slab and evaluation of structural stability under impact loads, based on Korean Highway Bridge Design Specifications and AASHTO LRFD Bridge Design Specification. According to analysis result, the maximum stress of edge beam and slab satisfies specification of allowable stress.