The College of Engineering and Computer Science

Permanent URI for this community

https://vinspace.online/handle/123456789/3

Education to create a generation of young leaders with the qualities, skills and knowledge needed to lead the development of science and technology, to increase productivity, and bring about radical innovation for the benefit of the society.

Browse

Now showing 1 - 20 of 65

A comprehensive imputation-based evaluation of tag SNP selection strategies
(2021) Dat Thanh Nguyen; Hieu Quang Dinh; Giang Minh Vu; Duong Thuy Nguyen; Nam Sy Vo
Regardless of the rapid development of sequencing technology, the single nucleotide polymorphism (SNP) array has been widely used for many large-scale genomic studies due to its cost-effectiveness. Recently, in parallel with the advancement in imputation strategies, several genotyping platforms for various species have been developed. Despite the importance of imputation accuracy in SNP array design, to the best of our knowledge, there are no systematic studies for evaluating tag SNP selection methods based on this metric. In this paper, using the leave-one-out cross-validation approach on the 1000 genome high-coverage dataset, we comprehensively evaluated four well-known tag SNP selection algorithms based on imputation accuracy. Our results showed that although all widely used methods for SNP array design can provide reasonable imputation accuracy, the pairwise linkage disequilibrium-based tag SNP selection algorithm achieves the best performance. Our pipelines for running evaluated algorithms and leave-one-out cross-validation are available for public use at https:/Igithub.comldatnguiTagSNP_evaluation.
A novel deep learning-based approach for sleep apnea detection using single-lead ECG signals
(2022) Tu Nguyen; Thao Nguyen; Khiem Le; Hieu Pham; Cuong Do
Sleep apnea (SA) is a type of sleep disorder characterized by snoring and chronic sleeplessness, which can lead to serious conditions such as high blood pressure, heart failure, and cardiomyopathy (enlargement of the muscle tissue of the heart). The electrocardiogram (ECG) plays a critical role in identifying SA since it might reveal abnormal cardiac activity. Recent research on ECG-based SA detection has focused on feature engineering techniques that extract specific characteristics from multiple-lead ECG signals and use them as classification model inputs. In this study, a novel method of feature extraction which based on the detection of S peaks is proposed to enhance the detection of adjacent SA segments using a single-lead ECG. In particular, ECG features collected from a single lead (V2) are used to identify SA episodes. On the extracted features, a CNN model is trained to detect SA. Experimental results demonstrate that the proposed method detects SA from single-lead ECG data is more accurate than existing state-of-the-art methods, with 91.13% classification accuracy, 92.58% sensitivity, and 88.75% specificity. Moreover, the further usage of features associated with the S peaks enhances the classification accuracy by 0.85%. Our findings indicate that the proposed machine learning system has the potential to be an effective method for detecting SA episodes.
A novel multi-view deep learning approach for BI-RADS and density assessment of mammograms
(2022) Huyen Nguyen; Sam Tran; Dung Nguyen; Hieu Pham; Ha Nguyen
Advanced deep learning (DL) algorithms may predict the patient’s risk of developing breast cancer based on the Breast Imaging Reporting and Data System (BI-RADS) and density standards. Recent studies have suggested that the combination of multi-view analysis improved the overall breast exam classification. In this paper, we propose a novel multi-view DL approach for BI-RADS and density assessment of mammograms. The proposed approach first deploys deep convolutional networks for feature extraction on each view separately. The extracted features are then stacked and fed into a Light Gradient Boosting Machine (LightGBM) classifier to predict BI-RADS and density scores. We conduct extensive experiments on both the internal mammography dataset and the public dataset Digital Database for Screening Mammography (DDSM). The experimental results demonstrate that the proposed approach outperforms the single-view classification approach on two benchmark datasets by huge F1-score margins (+5% on the internal dataset and +10% on the DDSM dataset). These results highlight the vital role of combining multi-view information to improve the performance of breast cancer risk prediction.
A Privacy-Preserving Framework for Surveillance Systems
(2020) Kok-Seng Wong; Tu Nguyen; Anuar Maratkhan; M. Fatih, Demirci
The ability to visually track people present in the scene is essential for any surveillance system. However, the widespread deployment and increased advancement of video surveillance systems have raised awareness of privacy to the public, i.e., human identity in the videos. The existing indoor surveillance systems allow people to be watched remotely and recorded continuously but do not prevent any party from viewing activities and collecting personal visual information of people in the videos. Because of this problem, we propose a privacy-preserving framework to provide each user (e.g., parents) with a personalized video where the user sees only selected target subjects (e.g., child, teacher, and intruder) while other faces are dynamically masked. The primary services in our framework consist of a video streaming service and a personalized service. The video streaming service is responsible for detecting, segmenting, recognizing, and masking face images of the human subjects in the video. Notably, it classifies human subjects into insider and outsider classes and then applies the de-identification (i.e., masking) to those in the insider class, including the target subjects. Subsequently, the personalized service receives the visual information (i.e., masked and unmasked faces) from the streaming service and processes it at the user’s mobile device. The output is then a personalized video for each user. For security reasons, we require surveillance videos stored in the cloud in an encrypted form. To ensure an individual remains anonymous in a group, we propose a dynamic masking approach to mask the human subjects in the video. Our framework can deliver both reliable visual privacy protection and video utility. For instance, users can have confidence that their target subjects are anonymized in other views. To utilize the personalized video, users can use analytics software installed on their mobile devices to analyze the activities of their target subjects.
A Quantitative Analysis of the Effect of Human Detection and Segmentation Quality in Person Re-identification Performance
(2019) Binh Nguyen; Quan Nguyen; Lan Le; Thuy Pham; Nam Pham
Person re-identification, a problem of person identity association across camera views at different locations and times, is the second step in two-steps system for automatic video surveillance: person detection, tracking and person reidentification. However, most of the reported person Re-ID methods deal with the human regions of interest (ROIs) which are extracted manually with well-aligned bounding boxes. They mainly focus on designing discriminative feature descriptors and relevant metric learning on these manually-cropped human ROIs. This paper aims to answer two questions: (1) Do human detection and segmentation affect the performance of person reidentification?; (2) How to overcome the effect of human detection and segmentation with the state-of-the-art method for person re-identification? To answer these two questions, quantitative evaluations have been performed for both single-shot and multishot scenarios of person re-identification. Different state-of-the-art methods for human detection and segmentation have been evaluated on two benchmark datasets (VIPeR and PRID2011). The obtained results allow us to give some suggestions for developing fully automatic video surveillance systems.
A Review of Forest Fire Combating Efforts, Challenges and Future Directions in Peninsular Malaysia, Sabah, and Sarawak
(2022-09-01) Yee Jian Chew; Shih Yin Ooi; Ying Han Pang; Kok-Seng Wong
Abstract: The land surface of Malaysia mostly constitutes forest cover. For decades, forest fires have been one of the nation’s most concerning environmental issues. With the advent of machine learning, many studies have been conducted to resolve forest fire issues. However, the findings and results have been very case-specific. Most experiments have focused on particular regions with independent methodology settings, which has hindered the ability of others to reproduce works. Another major challenge is lack of benchmark datasets in this domain, which has made benchmark comparisons almost impossible to conduct. To our best knowledge, no comprehensive review and analysis have been performed to streamline the research direction for forest fires in Malaysia. Hence, this paper was aimed to review all works aimed to combat forest fire issues in Malaysia from 1989 to 2021. With the proliferation of publicly accessible satellite data in recent years, a new direction of utilising big data platforms has been postulated. The merit of this approach is that the methodology and experiments can be reproduced. Thus, it is strongly believed that the findings and analysis shown in this paper will be useful as a baseline to propagate research in this domain.
A Subjective Study on User Perception Aspects in Virtual Reality
(2019-08-16) Huyen Tran; Nam Pham; Cuong Pham; Yong Ju Jung; Thang Truong
Three hundred and sixty degree video is becoming more and more popular on the Internet. By using a Head-Mounted Display, 360-degree video can render a Virtual Reality (VR) environment. However, it is still a big challenge to understand Quality of Experience (QoE) of 360-degree video since user experience during watching 360-degree video is a very complex phenomenon. In this paper, we aim to investigate four QoE aspects of 360-degree video, namely, perceptual quality, presence, cybersickness, and acceptability. In addition, four key QoE-affecting factors of encoding parameters, content motion, rendering device, and rendering mode are considered in our study. To the best of our knowledge, this is the first work that covers a large number of factors and QoE aspects of 360-degree video. In this study, a subjective experiment is conducted using 60 video versions generated from three original 360-degree videos. Based on statistical analysis of the obtained results, various findings on the impacts of the factors on the QoE aspects are provided. In particular, regarding the impacts of encoding parameters, it is found that the difference of QoE is negligible between video versions encoded at 4 K and 2.5 K resolutions. Also, it is suggested that 360-degree video should not be encoded at HD resolution or lower when watching in VR mode using Head Mounted Display. In addition, the bitrate for good QoE varies widely across different video contents. With respect to the content motion factor, its impact is statistically significant on the perceptual quality, presence, and cybersickness. In a comparison of two rendering device sets used in this study, there is no statistically significant difference found for the acceptability and cybersickness. However, the differences of the perceptual quality and presence are indicated to be statistically significant. Regarding the rendering mode, a comparison between VR and non-VR modes is also conducted. Although the non-VR mode always achieves higher perceptual quality scores and higher acceptability rates, more than half of the viewers prefer the VR mode to the non-VR mode when watching versions encoded at the resolutions of fHD or higher. By contrast, the non-VR mode is preferred at the HD resolution.
Active Temperature Compensation for MEMS Capacitive Sensor
(2021-09-01) Cuong Do; Ashwin A. Seshia
Temperature variations are one of the most crucial factors that need to be compensated for in MEMS sensors. Many traditional methodologies require an additional circuit to compensate for temperature. This work describes a new active temperature compensation method for MEMS capacitive strain sensors without additional circuits. The proposed method is based on a complement 2-D capacitive structure design. It consumes zero-power, which is essential toward the realization of a low-power temperature-compensated sensor in battery-powered or energy-harvesting applications. The gauge factor of the developed MEMS capacitive strain sensor is 7. The best result showing a capacitance variation of 1 ppm/◦C compared with 13 ppm/◦C on conventional design.
Adaptive Proxy Anchor loss for Deep Machine Learning
(2022) Nguyen Phan; Sen Tran; Huy Ta; Soan Duong; Chanh Nguyen; Trung Bui; Steven Truong
Deep metric learning (or simply called metric learning) uses the deep neural network to learn the representation of images, leading to widely used in many applications, e.g. image retrieval and face recognition. In the metric learning approaches, proxy anchor takes advantage of proxy-based and pair-based approaches to enable fast convergence time and robustness to noisy labels. However, in training the proxy anchor, selecting the hyperparameter margin is important to achieve a good performance. This selection requires expertise and is time-consuming. This paper proposes a novel method to learn the margin while training the proxy anchor approach adaptively. The proposed adaptive proxy anchor simplifies the hyperparameter tuning process while advancing the proxy anchor. We achieve state-of-the-art on three public datasets with a noticeably faster convergence time. Our code is available at https://github.com/tks1998/Adaptive-Proxy-Anchor
Adoption of IP Truncation in a Privacy-Based Decision Tree Pruning Design: A Case Study in Network Intrusion Detection System
(2022) Yee Jian Chew; Shih Yin Ooi; Kok-Seng Wong; Ying Han Pang; Nicholas Lee
A decision tree is a transparent model where the rules are visible and can represent the logic of classification. However, this structure might allow attackers to infer confidential information if the rules carry some sensitive information. Thus, a tree pruning methodology based on an IP truncation anonymisation scheme is proposed in this paper to prune the real IP addresses. However, the possible drawback of carelessly designed tree pruning might degrade the performance of the original tree as some information is intentionally opted out for the tree’s consideration. In this work, the 6-percentGureKDDCup’99, full-version GureKDDCup’99, UNSW-NB15, and CIDDS-001 datasets are used to evaluate the performance of the proposed pruning method. The results are also compared to the original unpruned tree model to observe its tolerance and trade-off. The tree model adopted in this work is the C4.5 tree. The findings from our empirical results are very encouraging and spell two main advantages: the sensitive IP addresses can be “pruned” (hidden) throughout the classification process to prevent any potential user profiling, and the number of nodes in the tree is tremendously reduced to make the rule interpretation possible while maintaining the classification accuracy.
An Efficient QoE-Aware HTTP Adaptive Streaming over Software Defined Networking
(2020) Thinh Pham; Dat Nguyen; Nam Pham; Thanh Nguyen; Hien Nguyen; Huong Truong
Due to the increase in video streaming traffic over the Internet, more innovative methods are in demand for improving both Quality of Experience (QoE) of users and the Quality of Service (QoS) of providers. In recent years, HTTP Adaptive Streaming (HAS) has received significant attention from both industry and academia based on its impacts on the enhancement of media streaming services. However, HAS alone cannot guarantee a seamless viewing experience, since this highly relies on the Network Operators’ infrastructure and evolving network conditions. Along with the development of future Internet infrastructure, Software-Defined Networking (SDN) has been researched and newly implemented as a promising solution in improving services of different Internet layers. In this paper, we present a novel architecture incorporating bitrate adaptation and dynamic route allocation. On the client side, the adaptation logic of VBR video streaming is built based on the MPEGDASH standard. On the network side, an SDN controller is implemented with several routing strategies on top of the OpenFlow protocol. Our experimental results show that the proposed solution enhances at least 38% up to 185% in terms of average bitrate in comparison with some existing solutions as well as achieves a better viewing experience than the traditional Internet.
An experimental method to characterize the relationship between aperture image and ray directions in microscope optics
(2020-10-21) Mai Tran; Oldenbourg, Rudolf
We propose a direct experimental method to calibrate the relationship between ray directions in object space and their positions in the aperture plane of a light field microscope. The calibration improves the interpretation of light field images, which contain information from both types of image planes, the field plane and the aperture plane of the ray path in the microscope. Our method is based on the diffraction of line gratings of known periodicities and provides accurate results with subpixel resolution. The method can be custom-tailored to most any optical configuration, including standard light microscopy setups, whenever correct mapping between ray parameters in the object/image plane and the aperture plane is needed.
An LSTM-based Approach for Overall Quality Prediction in HTTP Adaptive Streaming
(2019-04) Huyen Tran; Duc Nguyen; Nam Pham; Thang Truong; Duong Nguyen
HTTP Adaptive Streaming (HAS) has become a popular solution for multimedia delivery nowadays. In HAS, video quality is generally varying in each streaming session. Therefore, a key question in HTTP Adaptive Streaming is how to evaluate the overall quality of a streaming session. In this paper, we propose a machine-learning approach for overall quality prediction in HTTP Adaptive Streaming. In the proposed approach, each segment is represented by four features segment quality, stalling durations, content characteristics, and padding. The features are fed into a Long Short-Term Memory (LSTM) network that is capable of exploring temporal relations between segments. The overall quality of the streaming session is predicted from the outputs of the LSTM network using a linear regression module. Experiment results show that the proposed approach is effective in predicting the overall quality of streaming sessions. Also, it is found that our proposed approach outperforms four existing approaches.
Benchmarking full version of GureKDDCup, UNSW-NB15, and CIDDS-001 NIDS datasets using rolling-origin resampling
(2021) Yee Jian Chew; Nicholas Lee; Shih Yin Ooi; Kok-Seng Wong; Ying Han Pang
Network intrusion detection system (NIDS) is a system that analyses network traffic to flag malicious traffic or suspicious activities. Several recent NIDS datasets have been published, however, the lack of baseline experimental results on the full version of datasets had made it difficult for researchers to perform benchmarking. As the train-test distribution of the datasets has yet to be predefined by the creators, this further obstructs the researchers to compare the performance unbiasedly across each of the machine classifiers. Moreover, cross-validation resampling schemes have also been addressed in the literatures to be inappropriate in the domain of NIDS. Thus, rolling origin – a standard resampling technique that is also known as a common cross-validation scheme in the forecasting domain is employed to allocate the training and testing distributions. In this paper, rigorous experiments are conducted on the full version of the three recent NIDS datasets: GureKDDCup, UNSW-NB15, and CIDDS-001. While the datasets chosen might not be the latest available datasets, we have selected them as they include the essential IP addresses fields which are usually missing or removed due to some sort of privacy concerns. To deliver the baseline empirical results, 10 well-known classifiers from Weka are utilized.
Benchmarking saliency methods for chest X-ray interpretation
(2022-10-10) Adriel Saporta; Xiaotong Gui; Ashwin Agrawal; Anuj Pareek; Steven Truong; Chanh Nguyen; Doan Ngo; Jayne Seekins; Francis G. Blankenberg; Andrew Y. Ng; Matthew P. Lungren; Pranav Rajpurkar
Saliency methods, which produce heat maps that highlight the areas of the medical image that influence model prediction, are often presented to clinicians as an aid in diagnostic decision-making. However, rigorous investigation of the accuracy and reliability of these strategies is necessary before they are integrated into the clinical setting. In this work, we quantitatively evaluate seven saliency methods, including Grad-CAM, across multiple neural network architectures using two evaluation metrics. We establish the first human benchmark for chest X-ray segmentation in a multilabel classification set-up, and examine under what clinical conditions saliency maps might be more prone to failure in localizing important pathologies compared with a human expert benchmark. We find that (1) while Grad-CAM generally localized pathologies better than the other evaluated saliency methods, all seven performed significantly worse compared with the human benchmark, (2) the gap in localization performance between Grad-CAM and the human benchmark was the largest for pathologies that were smaller in size and had shapes that were more complex, and (3) model confidence was positively correlated with Grad-CAM localization performance. Our work demonstrates that several important limitations of saliency methods must be addressed before we can rely on them for deep learning explainability in medical imaging.
CapNeXt: Unifying capsule and resnext for medical image segmentation
(2022) Thanh Huynh; Chanh Nguyen; Khoa Nguyen; Trung Bui; Steven Truong
Capsule Network is a contemporary approach to image analysis that emphasizes part-whole relationships. However, its applications to segmentation tasks are limited due to training difficulties such as initialization and convergence. In this study, we propose a novel Capsule Network, called CapNeXt, that unifies Capsule and ResNeXt architectures for medical image segmentation. CapNeXt advances the existing capsule-based segmentation model by integrating optimization techniques from Convolutional Neural Networks (CNN) to make training much easier than other contemporary Capsule-based segmentation methods. Experimental results on two public datasets show that CapNeXt outperforms the CNNs and other Capsule architectures in 2D and 3D segmentation tasks by 1% of the Dice score. The code will be released on GitHub after being accepted.
Collaborative Curating for Discovery and Expansion of Visual Clusters
(2021) Dung Le; Hady W. Lauw
In many visually-oriented applications, users can select and group images that they find interesting into coherent clusters. For instance, we encounter these in the form of hashtags on Instagram, galleries on Flickr, or boards on Pinterest. The selection and coherence of such user-curated visual clusters arise from a user’s preference for a certain type of content as well as her own perception of which images are similar and thus belong to a cluster. We seek to model such curation behaviors towards supporting users in their future activities such as expanding existing clusters or discovering new clusters altogether. This paper proposes a framework, namely Collaborative Curating that jointly models the interrelated modalities of preference expression and similarity perception. Extensive experiments on real-world datasets of various categories from a visual curating platform show that the proposed framework significantly outperforms baselines focusing on either clustering behaviors or preferences alone.
ColorRL: Reinforced Coloring for End-to-End Instance Segmentation
(2021) Tuan Tran; Khoa Nguyen; Quan Tran; Won-Ki Jeong
Instance segmentation, the task of identifying and separating each individual object of interest in the image, is one of the actively studied research topics in computer vision. Although many feed-forward networks produce high-quality binary segmentation on different types of images, their final result heavily relies on the post-processing step, which separates instances from the binary mask. In comparison, the existing iterative methods extract a single object at a time using discriminative knowledge-based properties (e.g., shapes, boundaries, etc.) without relying on post-processing. However, they do not scale well with a large number of objects. To exploit the advantages of conventional sequential segmentation methods without impairing the scalability, we propose a novel iterative deep reinforcement learning agent that learns how to differentiate multiple objects in parallel. By constructing a relational graph between pixels, we design a reward function that encourages separating pixels of different objects and grouping pixels that belong to the same instance. We demonstrate that the proposed method can efficiently perform instance segmentation of many objects without heavy post-processing.
Constrained crystal growth during solidification of particles and splats in uniform droplet sprays
(2020) Yiannos Ioannou; Hiroki Fukuda; Claus Rebholz; Yiliang Liao; Teiichi Ando; Charalabos C. Doumanidis
Uniform droplet spraying (UDS) is a novel process used to produce ideally narrow (mono-size) distributions of molten metal droplets for various applications. The crystallite size is a primary determinant of mechanical properties in solidified alloy deposits and thus in need of predictive modeling. This project reports on employing UDS as a paradigm for solidification modeling of mono-size solid droplets in an oil bath, as well as planar and globular splats on a cooling substrate, for magnesium alloys AZ91D and Mg97ZnY2. The model combines a nucleation and dendrite fragmentation description from solidification theory, with a framework for constrained growth of crystalline domains confined by adjacent developing ones. The latter is based on differential attributes of the dynamic temperature field during solidification, derived from semi-analytical expressions for the simple droplet and splat geometries above. The modeling results are validated against measured domain size distributions on section micrographs and found to be within a − 10% to + 14% estimation error range. Further improvement of the model via numerical thermal descriptions for off-line material design and optimization in additive manufacturing is discussed, along with its use as a real-time structural observer for closed-loop control based on temperature measurements in UDS-based processes.
Cumulative Quality Modeling for HTTP Adaptive Streaming
(2021-04) Nam Pham; Huyen Tran; Truong, Cong Thang; Hobfeld, Tobias; Seufert, Michael
HTTP Adaptive Streaming has become the de facto choice for multimedia delivery. However, the quality of adaptive video streaming may fluctuate strongly during a session due to throughput fluctuations. So, it is important to evaluate the quality of a streaming session over time. In this article, we propose a model to estimate the cumulative quality for HTTP Adaptive Streaming. In the model, a sliding window of video segments is employed as the basic building block. Through statistical analysis using a subjective dataset, we identify four important components of the cumulative quality model, namely the minimum window quality, the last window quality, the maximum window quality, and the average window quality. Experiment results show that the proposed model achieves high prediction performance and outperforms related quality models. In addition, another advantage of the proposed model is its simplicity and effectiveness for deployment in real-time estimation. Our subjective dataset as well as the source code of the proposed model have been made publicly available at https://sites.google.com/site/huyenthithanhtran1191/cqmdatabase.

Browse

Browsing The College of Engineering and Computer Science by Title

Results Per Page

Sort Options