• Proceedings of the Korea Information Processing Society Conference
• /
• 2019.05a
• /
• pp.498-500
• /
• 2019

Point cloud is a flexible set of points that can provide a scalable geometric representation which can be applied in different computer graphic task. We propose a method based on EdgeConv and densely connected layers to aggregate the features for better classification. Our proposed approach shows significant performance improvement compared to the state-of-the-art deep neural network-based approaches.

• The Journal of Advanced Prosthodontics
• /
• v.9 no.3
• /
• pp.152-158
• /
• 2017

PURPOSE. The purpose of this study was to evaluate the influence of different coping thicknesses and veneer ceramic cooling rates on the failure load of zirconia-ceramic crowns. MATERIALS AND METHODS. Zirconia copings of two different thicknesses (0.5 mm or 1.5 mm; n=20 each) were fabricated from scanning 40 identical abutment models using a dental computer-aided design and computer-aided manufacturing system. Zirconia-ceramic crowns were completed by veneering feldspathic ceramics under different cooling rates (conventional or slow, n=20 each), resulting in 4 different groups (CONV05, SLOW05, CONV15, SLOW15; n=10 per group). Each crown was cemented on the abutment. 300,000 cycles of a 50-N load and thermocycling were applied on the crown, and then, a monotonic load was applied on each crown until failure. The mean failure loads were evaluated with two-way analysis of variance (P=.05). RESULTS. No cohesive or adhesive failure was observed after fatigue loading with thermocycling. Among the 4 groups, SLOW15 group (slow cooling and 1.5 mm chipping thickness) resulted in a significantly greater mean failure load than the other groups (P<.001). Coping fractures were only observed in SLOW15 group. CONCLUSION. The failure load of zirconia-ceramic crowns was significantly influenced by cooling rate as well as coping thickness. Under conventional cooling conditions, the mean failure load was not influenced by the coping thickness; however, under slow cooling conditions, the mean failure load was significantly influenced by the coping thickness.

• 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.

• Journal of the Korean Society for Marine Environment & Energy
• /
• v.1 no.2
• /
• pp.82-95
• /
• 1998

In this paper, two series of numerical simulations are performed using LS/DYNA3D: The first series of numerical simulations are collision events between a 310,000 DWT double hull VLCC (struck ship) and two 35,000 and 105,000 DWT tankers (striking ships). Collisions are assumed to occur at the middle of the VLCC with the striking ships moving at right angle to the YLCC centerline. The second ones, grounding accidents of two 40,000 DWT Conventional and Advanced Double Hull lanker bottom structures, CONV/PD328 and ADH/PD328 models. The overall objective of this study is to understand the structural failure and energy absorbing mechanisms during collision and grounding events for double hull tanker side and bottom structures, which lead to the initiation of inner shell rupture and cause the kinetic energy dissipation to bring the ship to a stop. These numerical simulations will contribute to the estimation of damage extents of collision and grounding accidents and the future improvements in lanker safety at the design stage.

• Journal of IKEEE
• /
• v.24 no.1
• /
• pp.225-231
• /
• 2020

In this paper, we propose a method to accelerate BWN based on FPGA with limited resources for embedded system. Because of the limited number of logic elements available, a single computing unit capable of handling Conv-layer, FC-layer of various sizes must be designed and reused. Also, if the input feature map can not be parallel processed at one time, the output must be calculated by reading the inputs several times. Since the number of available BRAM modules is limited, the number of data bits in the BWN accelerator must be minimized. The image classification processing time of the BWN accelerator is superior when compared with a embedded CPU and is faster than a desktop PC and 50% slower than a GPU system. Since the BWN accelerator uses a slow clock of 50MHz, it can be seen that the BWN accelerator is advantageous in performance versus power.

• Journal of the Institute of Electronics Engineers of Korea SD
• /
• v.46 no.3
• /
• pp.75-85
• /
• 2009

This work proposes a 13b 100MS/s 0.13um CMOS ADC for 3G communication systems such as two-carrier W-CDMA applications simultaneously requiring high resolution, low power, and small size at high speed. The proposed ADC employs a four-step pipeline architecture to optimize power consumption and chip area at the target resolution and sampling rate. Area-efficient high-speed high-resolution gate-bootstrapping circuits are implemented at the sampling switches of the input SHA to maintain signal linearity over the Nyquist rate even at a 1.0V supply operation. The cascode compensation technique on a low-impedance path implemented in the two-stage amplifiers of the SHA and MDAC simultaneously achieves the required operation speed and phase margin with more reduced power consumption than the Miller compensation technique. Low-glitch dynamic latches in sub-ranging flash ADCs reduce kickback-noise referred to the differential input stage of the comparator by isolating the input stage from output nodes to improve system accuracy. The proposed low-noise current and voltage references based on triple negative T.C. circuits are employed on chip with optional off-chip reference voltages. The prototype ADC in a 0.13um 1P8M CMOS technology demonstrates the measured DNL and INL within 0.70LSB and 1.79LSB, respectively. The ADC shows a maximum SNDR of 64.5dB and a maximum SFDR of 78.0dB at 100MS/s, respectively. The ABC with an active die area of $1.22mm^2$ consumes 42.0mW at 100MS/s and a 1.2V supply, corresponding to a FOM of 0.31pJ/conv-step.

• Journal of Intelligence and Information Systems
• /
• v.29 no.1
• /
• pp.249-263
• /
• 2023

Collaborative Filtering, a representative recommendation system methodology, consists of two approaches: neighbor methods and latent factor models. Among these, the latent factor model using matrix factorization decomposes the user-item interaction matrix into two lower-dimensional rectangular matrices, predicting the item's rating through the product of these matrices. Due to the factor vectors inferred from rating patterns capturing user and item characteristics, this method is superior in scalability, accuracy, and flexibility compared to neighbor-based methods. However, it has a fundamental drawback: the need to reflect the diversity of preferences of different individuals for items with no ratings. This limitation leads to repetitive and inaccurate recommendations. The Adaptive Deep Latent Factor Model (ADLFM) was developed to address this issue. This model adaptively learns the preferences for each item by using the item description, which provides a detailed summary and explanation of the item. ADLFM takes in item description as input, calculates latent vectors of the user and item, and presents a method that can reflect personal diversity using an attention score. However, due to the requirement of a dataset that includes item descriptions, the domain that can apply ADLFM is limited, resulting in generalization limitations. This study proposes a Generalized Adaptive Deep Latent Factor Recommendation Model, G-ADLFRM, to improve the limitations of ADLFM. Firstly, we use item ID, commonly used in recommendation systems, as input instead of the item description. Additionally, we apply improved deep learning model structures such as Self-Attention, Multi-head Attention, and Multi-Conv1D. We conducted experiments on various datasets with input and model structure changes. The results showed that when only the input was changed, MAE increased slightly compared to ADLFM due to accompanying information loss, resulting in decreased recommendation performance. However, the average learning speed per epoch significantly improved as the amount of information to be processed decreased. When both the input and the model structure were changed, the best-performing Multi-Conv1d structure showed similar performance to ADLFM, sufficiently counteracting the information loss caused by the input change. We conclude that G-ADLFRM is a new, lightweight, and generalizable model that maintains the performance of the existing ADLFM while enabling fast learning and inference.

• International journal of advanced smart convergence
• /
• v.8 no.4
• /
• pp.75-81
• /
• 2019

Though recent advancement of deep learning methods have provided satisfactory results from large data domain, somehow yield poor performance on few-shot classification tasks. In order to train a model with strong performance, i.e. deep convolutional neural network, it depends heavily on huge dataset and the labeled classes of the dataset can be extremely humongous. The cost of human annotation and scarcity of the data among the classes have drastically limited the capability of current image classification model. On the contrary, humans are excellent in terms of learning or recognizing new unseen classes with merely small set of labeled examples. Few-shot learning aims to train a classification model with limited labeled samples to recognize new classes that have neverseen during training process. In this paper, we increase the backbone depth of the embedding network in orderto learn the variation between the intra-class. By increasing the network depth of the embedding module, we are able to achieve competitive performance due to the minimized intra-class variation.

• Proceedings of the Korean Society of Computer Information Conference
• /
• 2016.07a
• /
• pp.99-100
• /
• 2016

본 논문에서는 근골격계 질환 중에서 가장 흔히 나타나는 어깨의 질환에 대하여 살펴보도록 한다. 특히 어깨를 옆으로 반드시 폈다가 안으로 오므리는 수평 벌림과 수평 모음에 대한 측정 방법을 살펴본다. 여기서 일정한 임계값보다 크게 나타나는 값인 통증 유발에는 반드시 도수 치료가 필요하다고 본다.

• Proceedings of the Korean Society of Computer Information Conference
• /
• 2016.07a
• /
• pp.97-98
• /
• 2016

본 논문에서는 흔히 나타나는 어깨 질환의 치료에서 팔 부분에 대하여 알아본다. 특히 팔을 반듯이 위로 올렸을 때 통증이 발생하는 어깨 질환에 대한 도수치료법(Manual Therapy)을 제시한다. 실험에서는 도수진단(Manual Diagnosis)을 위하여 팔을 반듯이 위로 올릴 때 팔의 각도를 입력하고, 이상이 발생할 시 진단 및 치료법까지 시스템에서 제시하도록 한다.