1,720,963 research outputs found
Artificial Intelligence for Skin Lesion Analysis based on Computer Vision and Deep Learning
Get PDFSkin lesions appear in various sizes and forms and can be localised in one place or spread across the whole body due to different conditions. Dermatologists typically undertake physical examinations to diagnose skin lesions. However, this task costs time and requires excessive effort and can be inconsistent. Depending on the type of lesion and whether or not malignancy is present, additional diagnostic testing, such as imaging or biopsy, may be needed. Computer-aided diagnosis (CAD) systems, using clinical and dermoscopic images, could provide a quantitative assessment tool to help clinicians identify skin lesions and evaluate their severity. The recent progress in computer vision and deep learning has encouraged researchers to harness medical imaging data to develop powerful tools which could provide better diagnosis, treatment and prediction of skin conditions.
By leveraging artificial intelligence techniques, including computer vision and deep learning, this work introduces intelligent computerised approaches using dermoscopic and clinical images to analyse and identify two types of skin lesions producing enhanced medical information. This thesis designed, realised, and evaluated the benefit of features learned automatically from images through the stacked layers of convolution filters in the convolutional neural network (CNN) models. The final objective of conducting the research in this thesis is to benefit patients with skin lesion condition assessment and skin cancer identification without adding to the already high medical costs. An automated regression-based method has been developed in this thesis for acne counting and severity grading from clinical facial images. In addition to the acne lesions, another type of skin lesion has been considered, represented by melanoma-related lesions. Two pipelines have been presented in this thesis to identify melanoma lesions. The first framework benchmarks and evaluates several CNN models for melanoma and non- melanoma classification from only dermoscopic images. While the second developed model for melanoma detection integrates the seven-point checklist scheme with CNN using both clinical and dermoscopic images.
The experimental results of the work presented in this thesis manifest improved/ competitive performance compared to the state-of-the-art skin analysis methods using several evaluation metrics. The findings of the developed approaches demonstrated effective analysis of skin lesions with high accuracy, reducing the risk of misdiagnosis, and providing a more efficient means of detecting melanoma and automated acne lesion severity grading. Additionally, the application of computational intelligence allows for cost savings by reducing the need for manual analysis and enabling the automation of grading support, resulting in a more reliable and consistent process. Overall, the new automated methods based on computational intelligence demonstrate the benefits of developing computer vision and deep learning techniques for skin lesion analysis towards early skin cancer identification and cost-effective and robust grading support
Improved Performance of Secured VoIP Via Enhanced Blowfish Encryption Algorithm
Get PDFBoth the development and the integration of efficient network, open source technology, and Voice over Internet Protocol (VoIP) applications have been increasingly important and gained quick popularity due to new rapidly emerging IP-based network technology. Nonetheless, security and privacy concerns have emerged as issues that need to be addressed. The privacy process ensures that encryption and decryption methods protect the data from being alternate and intercept, a privacy VoIP call will contribute to private and confidential conversation purposes such as telebanking, telepsychiatry, health, safety issues and many more. Hence, this study had quantified VoIP performance and voice quality under security implementation with the technique of IPSec and the enhancement of the Blowfish encryption algorithm. In fact, the primary objective of this study is to improve the performance of Blowfish encryption algorithm. The proposed algorithm was tested with varying network topologies and a variety of audio codecs, which contributed to the impact upon VoIP network.
A network testbed with seven experiments and network configurations had been set up in two labs to determine its effects on network performance. Besides, an experimental work using OPNET simulations under 54 experiments of network scenarios were compared with the network testbed for validation and verification purposes. Next, an enhanced Blowfish algorithm for VoIP services had been designed and executed throughout this research. From the stance of VoIP session and services performance, the redesign of the Blowfish algorithm displayed several significant effects that improved both the performance of VoIP network and the quality of voice. This finding indicates some available opportunities that could enhance encrypted algorithm, data privacy, and integrity; where the balance between Quality of Services (QoS) and security techniques can be applied to boost network throughput, performance, and voice quality of existing VoIP services. With that, this study had executed and contributed to a threefold aspect, which refers to the redesign of the Blowfish algorithm that could minimize computational resources. In addition, the VoIP network performance was analysed and compared in terms of end-to-end delay, jitter, packet loss, and finally, sought improvement for voice quality in VoIP services, as well as the effect of the designed enhanced Blowfish algorithm upon voice quality, which had been quantified by using a variety of voice codecs
Artificial Intelligence Applied to Facial Image Analysis and Feature Measurement
Get PDFBeauty has always played an essential part in society, influencing both everyday human interactions and more significant aspects such as mate selection. The continued and expanding use of beauty products by women and, increasingly, men worldwide has prompted and motivated several companies to develop platforms that effectively integrate into the beauty and cosmetics sector. They attempt to improve the customer experience by combining data with personalisation. Global cosmetics spending is worth billions of dollars, and most of it is wasted on unsuitable or incompatible products. This enables artificial intelligence to alter the rules using computer vision and deep learning approaches, allowing customers to be completely satisfied. With the advanced feature extraction in deep learning, especially convolutional neural networks, automatic facial feature analysis from images for the sake of beauty and beautification has become an emerging subject of study. Scholars studying facial aesthetics have recently made breakthroughs in the areas of facial shape beautification and beauty prediction. In the cosmetics sector, a new line of recommendation system research has arisen. Users benefit from recommendation systems since these systems help them narrow down their options. This thesis has laid the groundwork for a recommendation system related to beautification purposes through hairstyle and eyelashes leveraging artificial intelligence techniques. One of the most potent descriptors for attribution of personality is facial attributes. Various types of facial attributes are extracted in this thesis, including geometrical, automatic and hand-crafted features. The extracted attributes provide rich information for the recommendation system to produce the final outcome. The coexistence of external effects on the faces, like makeup or retouching, could disguise facial features. This might result in degradation in the performance of facial feature extraction and subsequently in the recommendation system. Thus, three methods are further developed to detect the faces wearing the makeup before passing the images into the recommendation system. This would help to provide more reliable and accurate feature extraction and suggest more suitable recommendation results. This thesis also presents a method for segmenting the facial region with the goal of extending the developed recommendation system by incorporating a synthesised hairstyle virtually on the facial region, thereby harnessing the recommended hairstyle generated by our developed system. Hence, the work presented in this thesis shows the benefits of implementing computational intelligence methods in the beauty and cosmetics sector. It also demonstrates that computational intelligence techniques have redefined the notion of beauty and how the consumer communicates with these emerging intelligent facilities that bring solutions to our fingertips
Detection, Prediction and Modelling of Mental Fatigue in Naturalistic Environment
Get PDFOperator mental fatigue in workplace can result in serious mistakes which have dangerous and life-threatening consequences. Fatigue assessment and prediction are, therefore, considered critical safety requirements that cut across modes and operations of numerous high-risk environments and industries such as nuclear and transportation. However, robust, accurate and timely assessment of fatigue (or alertness) is still a challenging task for many reasons. The majority of operator fatigue studies are still being carried out in simulation environments, overlooking operator's naturalistic behaviour and fatigue growth. Moreover, most of the available systems rely on using a single fatigue-related data source, which is clearly a major drawback that affects operation, performance, accuracy and reliability of the system in case this source fails. With multi-data sources in an integrated system, the system might stop working in the event of losing one or more data sources or at least becomes inaccurate or unreliable. Furthermore, paying no attention to human individual differences working as an operator in mission-critical jobs related to fatigue growth and in response to fatigue deleterious effect is another serious issue with the current fatigue assessment and prediction systems.
The research work presented in this thesis proposes a novel fatigue assessment approach, which addresses the aforementioned issues with fatigue detection and prediction system. This is achieved by developing and realising algorithms based on data collected from participants in naturalistic environments. Numerous experiments have been conducted to cover a wide range of fatigue-related tasks which are broadly grouped into two categories: biological and behavioural (performance) experiments. The biological-based experiments employ various data types such as heart rate, skin temperature, skin conductance and heart rate variability. These fatigue-related data types are used to build the proposed fatigue detection system, and the obtained results have demonstrated high accuracy and reliability (94.5% accuracy in naturalistic environments). The behavioural-based category includes two experiments: keyboard typing and driving task. The typing experiments have been carried out using computer keyboard and smartphone virtual keyboard, and have confirmed enhanced operator fatigue detection accuracy (94%). The driving experiments were conducted in naturalistic driving environments, and the used algorithms have demonstrated a new framework for driver fatigue detection using smartphone inertial sensors based on a novel vehicle heading algorithm.
A prototype system was designed and built with a modular structure so as to allow the addition of multiple fatigue-related biological and behavioural sources. This modular structure was tested under different situations that involve losing one or more sources. In addition, the circadian rhythm, which is a main input to fatigue/alert regulators, was customised for each operator and modelled based on biological data collected from wearable devices. The constructed model captures individual differences of operators, which is a challenge in current systems. Such multi-source, modular and non-intrusive approach for fatigue/alertness assessment and prediction is expected to be of superior performance, low-cost and favourable by users compared to existing systems. Furthermore, it addresses other challenges of current fatigue systems by carrying out fatigue assessment in naturalistic environments and considering operator individual differences in response to fatigue. In addition, the modular structure of the proposed system helps improving robustness and accuracy against losing one or more input sources (accuracy for 4 sources: 91%, 3 sources: 87%, 2 sources: 77%). Following the proposed approach will definitely enhance the reliability of fatigue assessment systems, improve operator safety, productivity and reduce financial fatigue impacts. Moreover, the proposed system has proven to be non-intrusive in nature and of low implementation cost. The results obtained after testing the proposed system have been very promising to support the aforementioned benefits
Hybrid Learning Using Canvas LMS
Get PDFHybrid learning refers to the learning style where online components are used to replace some face-to-face elements of the course. In the current era, where COVID-19 pandemic has highly impacted higher education, online and remote forms of learning have become critical success factors to deliver engaging and rich teaching and learning experience to the students. With the partial return of face-to-face interaction with the students this year after easing restrictions, universities have no choice but to offer hybrid learning experience. In the journey towards this type of learning, a transition in both pedagogy and vehicle (tools) is inevitable. Hybrid learning pedagogy has been in literature for many years now and many institutions worldwide have enough experience to run courses and programs as a hybrid model. For the vehicle, a number of tools are necessary to facilitate delivery, but the most important tool is obviously the learning management system (LMS). Canvas LMS is now considered one of the most commonly used electronic learning systems, offering a large number of features and options to make teaching and learning easier and effective for both teachers and students. In this paper, two hybrid learning models are proposed. An example of implementing one of the two models using Canvas LMS and other supporting tools is provided. Anecdotal student feedback has shown that the students were highly engaged and their experience has been improved as a result of the hybrid delivery format.</jats:p
Technology-Enhanced Learning and Teaching in COVID-19 Era: Challenges and Recommendations
Get PDFTechnology-enhanced learning and teaching methods have been in literature and for many years now. Many educational institutes all over the world have been using these methods to deliver their programs and degrees. Nevertheless, some institutes are not very keen on using technology in some disciplines, and deliver their programs in a traditional way for a number of reasons, especially if these have been successful and well-attended (i.e. popular) by students. In the current era, where COVID-19 pandemic has disrupted every corner of our life including higher education, technology has become a critical success factor to reduce the negative impact of this pandemic. Accordingly, it is now no longer an option to opt out from using technology in learning and teaching. This doesn’t just refer to providing (dumping) contents to students digitally, but to facilitate learning and deliver engaging and highly interactive experience to compensate for lack of face-to-face interaction between the students and their teachers and also amongst the students themselves. The use of technology in education due to COVID-19 pandemic, however, has confronted by a number of challenges. In some cases, the focus was shifted to the contents (documents, videos…etc.) rather than interactivity and student engagement. Furthermore, the students were highly overwhelmed with contents in a short period of time, which has caused anxiety, dissatisfaction and performance issues. In this paper, examples of teaching methods based on the use of technology that are employed during the lockdown period are provided. Moreover, a number of subsequent challenges due to current situation are discussed, and recommendations for implementation and best practice are shared. Also a proposal for a flipped delivery model to move forward is provided and discussed. Anecdotal student feedback has shown that the used methods and techniques were really helpful and have boosted student learning and enthusiasm in this difficult time. </jats:p
Automatic detection, sizing and characterisation of weld defects using ultrasonic time-of-flight diffraction
No full textUltrasonic time-of-flight diffraction (TOFD) is known as a reliable non-destructive
testing technique for weld inspection in steel structures, providing accurate aw
positioning and sizing. Despite all its good features, TOFD data interpretation
and reporting are still performed manually by skilled inspectors and interpretation
software operators. This is a cumbersome and error-prone process, leading to
inevitable delay and inconsistency. The quality of the collected TOFD data is
another issue that may introduce a host of error to the overall interpretation
process. Manual interpretation focuses only on the compression waves portion
of the collected TOFD data and overlooks the mode-converted waves region and
considers it redundant. This region may provide useful and accurate aw sizing
and classification information when there is uncertainty or ambiguity due to the
nature of the collected data or the type of aw, and can reduce the number of
supplementary (parallel) B-scans by utilising the (longitudinal) D-scans only. The
automation of data processing in TOFD is required to minimise time and error
and towards building a comprehensive computer-aided TOFD interpretation tool
that can aid human operators.
This project aims at proposing interpretation algorithms to size and characterise
flaws automatically and accurately using data acquired from D-scans only. In order
to achieve this, a number of novel data manipulation and processing techniques
have been specifically developed and adapted to expose the information in the
mode-converted waves region. In addition, several multi-resolution approaches
employing the Wavelet transform and texture analysis have been used in aw
detection and for de-noising and enhancing quality of the collected data.
Performance of the developed algorithms and the results of their application have
been promising in terms of speed, accuracy and consistency when compared to
human interpretation by an expert operator, using the compression waves portion
of the acquired data. This is expected to revolutionise the TOFD data interpretation and be in favour of a real-time processing of large volumes of data. It is highly anticipated that the research findings of this project will increase significantly the reliance on D-scans to obtain high sizing accuracy without the need to
perform further B-scans. The overall inspection and interpretation time and cost
will therefore be reduced significantly
Going Beyond Counting First Authors in Author Co-citation Analysis
Get PDFThe present study examines one of the fundamental aspects of author co-citation analysis (ACA) - the way co-citation
counts are defined. Co-citation counting provides the data on which all subsequent statistical analyses and mappings
are based, and we compare ACA results based on two different types of co-citation counting - the traditional type that
only counts the first one among a cited work's authors on the one hand and a non-traditional type that takes into
account the first 5 authors of a cited work on the other hand. Results indicate that the picture produced through this non-traditional author co-citation counting contains more coherent author groups and is therefore considerably clearer. However, this picture represents fewer specialties in the research field being studied than that produced through the traditional first-author co-citation counting when the same number of top-ranked authors is selected and analyzed. Reasons for these effects are discussed
Variations on the Author
Get PDF“Variations on the Author” discusses two of Eduardo Coutinho’s recent films (Um Dia na Vida, from 2010, and Últimas Conversas, posthumously released in 2015) and their contribution to the general question of documentary authorship. The director’s filmography is characterized by a consistent yet self-effacing form of authorial self-inscription: Coutinho often features as an interviewer that rather than express opinions propels discourses; an interviewer that is good at listening. This mode of self-inscription characterizes him as an author who is not expressive but who is nonetheless markedly present on the screen. In Um Dia na Vida, however, Coutinho is completely absent form the image, while Últimas Conversas, on the contrary, includes a confessional prologue that moves the director from the margins to the center of his films. This article examines the ways in which these works stand out in the filmography of a director who offers new insights into the notion of cinematic authorship
- …
