• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.17 no.9
• /
• pp.2505-2528
• /
• 2023

The field of facial expression recognition (FER) has been actively researched to improve human-computer interaction. In recent years, deep learning techniques have gained popularity for addressing FER, with numerous studies proposing end-to-end frameworks that stack or widen significant convolutional neural network layers. While this has led to improved performance, it has also resulted in larger model sizes and longer inference times. To overcome this challenge, our work introduces a novel lightweight model architecture. The architecture incorporates three key factors: Depth-wise Separable Convolution, Residual Block, and Attention Modules. By doing so, we aim to strike a balance between model size, inference speed, and accuracy in FER tasks. Through extensive experimentation on popular benchmark FER datasets, our proposed method has demonstrated promising results. Notably, it stands out due to its substantial reduction in parameter count and faster inference time, while maintaining accuracy levels comparable to other lightweight models discussed in the existing literature.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.11 no.3
• /
• pp.1700-1721
• /
• 2017

The problem of blocking artifacts is very common in block-based image and video compression, especially at very low bit rates. In this paper, we propose a post-processing method for JPEG-coded image deblocking via sparse representation and adaptive residual threshold. This method includes three steps. First, we obtain the dictionary by online dictionary learning and the compressed images. The dictionary is then modified by the histogram of oriented gradient (HOG) feature descriptor and K-means cluster. Second, an adaptive residual threshold for orthogonal matching pursuit (OMP) is proposed and used for sparse coding by combining blind image blocking assessment. At last, to take advantage of human visual system (HVS), the edge regions of the obtained deblocked image can be further modified by the edge regions of the compressed image. The experimental results show that our proposed method can keep the image more texture and edge information while reducing the image blocking artifacts.

• Composites Research
• /
• v.28 no.4
• /
• pp.182-190
• /
• 2015

The phenomenological evolution laws of damage can be defined either based on residual life or residual strength. The failure of a specimen can be defined immediately after or before fracture. The former is called in this paper by "failure defined by approach I" and the latter "failure defined by approach II." Usually at failure there is a discontinuity of loading variables and, because of this, damage at failure is discontinuous. Therefore the values of damage at failure by two different approaches are not the same. Based on this idea the sequence effects of the phenomenological evolution law of damage given by $dD/dN=g(D)f({\Phi})$ were studied. Thin-walled graphite/epoxy tubes consisting of four of $[{\pm}45]_s$ laminates were used for the experimental study of sequence effects and the effects of mean stress on fatigue life. It was found that the sequence effects in two step uniaxial fatigue for $[{\pm}45]_s$ graphite/epoxy tubular specimen showed that a high-low block loading sequence was less damaging than a low-high one.

• The Transactions of The Korean Institute of Electrical Engineers
• /
• v.68 no.1
• /
• pp.102-107
• /
• 2019

This paper proposes a method to improve the mass spectral detection performance of the residual gas analyzer. By improving the mode estimation method for setting the threshold value and improving the additive noise elimination method, it is possible to detect mass spectrums having low peak values of the threshold level difficult to distinguish from noise. Ion signal blocks for each mass index with noise removed by the improved method are effective for eliminating invalid ion signals based on the linear and quadratic fittings. The mass spectrum can be obtained from the quadratic fitted curves for the reconstructed ion signal block using only the valid ion signals. In addition, the resolution of the mass spectrum can be improved by correcting the error caused by the shift of the spectral peak position. To verify the performance of the proposed method, computer simulations were performed using real ion signals obtained from the residual gas analysis system under development. The simulation results show that the proposed method is valid.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.16 no.1
• /
• pp.1-15
• /
• 2022

The MobileNetV3 is specially designed for mobile devices with limited memory and computing power. To reduce the network parameters and improve the network inference speed, a new lightweight network is proposed based on MobileNetV3. Firstly, to reduce the computation of residual blocks, a partial residual structure is designed by dividing the input feature maps into two parts. The designed partial residual structure is used to replace the residual block in MobileNetV3. Secondly, a dual-path feature extraction structure is designed to further reduce the computation of MobileNetV3. Different convolution kernel sizes are used in the two paths to extract feature maps with different sizes. Besides, a transition layer is also designed for fusing features to reduce the influence of the new structure on accuracy. The CIFAR-100 dataset and Image Net dataset are used to test the performance of the proposed partial residual structure. The ResNet based on the proposed partial residual structure has smaller parameters and FLOPs than the original ResNet. The performance of improved MobileNetV3 is tested on CIFAR-10, CIFAR-100 and ImageNet image classification task dataset. Comparing MobileNetV3, GhostNet and MobileNetV2, the improved MobileNetV3 has smaller parameters and FLOPs. Besides, the improved MobileNetV3 is also tested on CPU and Raspberry Pi. It is faster than other networks

• The Korean Journal of Pain
• /
• v.27 no.2
• /
• pp.125-132
• /
• 2014

Background: To improve residual pain management in acute cervical radiculopathy treated with NSAIDs, the feasibility of early and repeated low-dose interscalene brachial plexus block (IS-BPB) needs to be assessed. Methods: This was a prospective study on patients receiving NSAIDs (loxoprofen) for cervical radiculopathy of ${\leq}2$-week onset. Pain was assessed using the visual analogue scale (VAS). A low-dose ultrasonography (USG)-guided IS-BPB (dexamethasone [1.65 mg; 0.5 ml] and mepivacaine [1%; 3.0 ml]) was performed at baseline and weekly thereafter for 4 weeks in an outpatient setting for the intervention group. All patients were evaluated using a visual satisfaction score (VSS) at week 4. Patients with baseline VAS scores < 70 (mild to moderate pain; MM group) and ${\geq}70$ (severe pain; SE group) were compared to the controls receiving NSAIDs. Results: A total of 316 IS-BPBs were performed in the intervention group. There was a significant difference in the decline in the VAS from week 0 to week 3 in the MM and SE groups (P < 0.05); however, from week 3 to week 4, the therapeutic effect exhibited no significant difference. Thirteen patients at week 2 (15.5%; MM: 27.7%; SE: 0%), 43 at week 3 (51.2%; MM: 83.0%; SE: 10.8%), and 47 at week 4 (56.0%; MM: 85.1%; SE: 18.9%) achieved a VAS score of ${\leq}20$. Patient satisfaction was high, and the decrease in VAS scores in both groups was significant (P < 0.05) compared to the controls. Conclusions: Weekly, low-dose, USG-guided IS-BPB can be implemented for early pain relief in acute cervical radiculopathy, with high patient satisfaction.

• The Korean Journal of Applied Statistics
• /
• v.34 no.6
• /
• pp.979-993
• /
• 2021

Acoustic scene classification (ASC) categorizes an audio file based on the environment in which it has been recorded. This has long been studied in the detection and classification of acoustic scenes and events (DCASE). In this study, we considered the problem that ASC faces in real-world applications that the model used should have low-complexity. We compared several models that apply light-weight techniques. First, a base CNN model was proposed using log mel-spectrogram, deltas, and delta-deltas features. Second, depthwise separable convolution, linear bottleneck inverted residual block was applied to the convolutional layer, and Quantization was applied to the models to develop a low-complexity model. The model considering low-complexity was similar or slightly inferior to the performance of the base model, but the model size was significantly reduced from 503 KB to 42.76 KB.

• Journal of Broadcast Engineering
• /
• v.11 no.2 s.31
• /
• pp.153-163
• /
• 2006

We propose a new video compression method based on MCTF(motion compensated temporal filtering) with constant quality. SPECK is an efficient image compression coding method of encoding DWT coefficients. Especially SPECK method is very efficient for coding the motion compensated residual image which usually has larger amounts of high frequency components than the natural images. And proposed multi block size hierarchical motion estimation technique is more efficient than classical block matching algorithm with fixed block size both in estimation precision and operation costs. Proposed video method based on MCTF video compression can also support multi-frame rate decoding with reasonable complexity. Simulation results showed that proposed method outperforms H.263 video compression standard.

• Journal of Broadcast Engineering
• /
• v.15 no.5
• /
• pp.653-664
• /
• 2010

Generally existing video codec employs entropy coding to deal with residual signals with considering temporal and spatial properties. Scalable Video Coding(SVC) which is extension of H.264/AVC has three technical concepts for removing redundancies between inter-layers. In spite of using novel prediction method between inter-layers in SVC, it is still using same entropy coding method to residual signals. According to the studies, the residual obtained by inter-layer prediction technique has different features of residual signal acquired by spatial or temporal prediction technique. In this paper, we propose an efficient entropy coding method which codes the residual signal obtained by inter-layer prediction with regarding its features adequately. We re-designed the Coded Block Pattern(CBP) table suitably for inter-layer texture prediction. The experiments show that the proposed method can further reduce the BD-Bitrate up to average 2.20% in 4CIF and 1.14% in CIF resolution compared to the existing JSVM 9.18.

• Journal of Wetlands Research
• /
• v.22 no.1
• /
• pp.1-7
• /
• 2020

Non-point pollution adversely affects the water system and its influence is increasing. In order to manage such nonpoint source pollution, the government has conducted studies on LID (Low Impact Development) facilities and various efficiency evaluations. In this study, the actual installed permeable block facility among the various LID facilities was analyzed the effluent reduction rate, the residual rainfall analysis, the runoff duration time and the reduction rate of the maximum inflow and outflow for the rainfall runoff control and the results were compared the other facilities. The analysis results show that the reduction efficiency is high in order of impermeable block, filter type permeable block, and clearance type permeable block, and the graph showing the relationship between the rainfall intensity and the runoff duration time is presented. This graph can be helpful in the design of facilities such as the facility capacity selection according to the reproduction period of the permeable block facility similar to this.