Deep Learning Approach to Image Deblurring and Image Super-Resolution using DeblurGAN and SRGAN
- Authors: Kuhlane, Luxolo L , Brown, Dane L , Connan, James , Boby, Alden , Marais, Marc
- Date: 2022
- Subjects: To be catalogued
- Language: English
- Type: text , article
- Identifier: http://hdl.handle.net/10962/465157 , vital:76578 , xlink:href="https://www.researchgate.net/profile/Luxolo-Kuhlane/publication/363257796_Deep_Learning_Approach_to_Image_Deblurring_and_Image_Super-Resolution_using_DeblurGAN_and_SRGAN/links/6313b5a01ddd44702131b3df/Deep-Learning-Approach-to-Image-Deblurring-and-Image-Super-Resolution-using-DeblurGAN-and-SRGAN.pdf"
- Description: Deblurring is the task of restoring a blurred image to a sharp one, retrieving the information lost due to the blur of an image. Image deblurring and super-resolution, as representative image restoration problems, have been studied for a decade. Due to their wide range of applications, numerous techniques have been proposed to tackle these problems, inspiring innovations for better performance. Deep learning has become a robust framework for many image processing tasks, including restoration. In particular, generative adversarial networks (GANs), proposed by [1], have demonstrated remarkable performances in generating plausible images. However, training GANs for image restoration is a non-trivial task. This research investigates optimization schemes for GANs that improve image quality by providing meaningful training objective functions. In this paper we use a DeblurGAN and Super-Resolution Generative Adversarial Network (SRGAN) on the chosen dataset.
- Full Text:
- Date Issued: 2022
Investigating signer-independent sign language recognition on the lsa64 dataset
- Authors: Marais, Marc , Brown, Dane L , Connan, James , Boby, Alden , Kuhlane, Luxolo L
- Date: 2022
- Subjects: To be catalogued
- Language: English
- Type: text , article
- Identifier: http://hdl.handle.net/10962/465179 , vital:76580 , xlink:href="https://www.researchgate.net/profile/Marc-Marais/publication/363174384_Investigating_Signer-Independ-ent_Sign_Language_Recognition_on_the_LSA64_Dataset/links/63108c7d5eed5e4bd138680f/Investigating-Signer-Independent-Sign-Language-Recognition-on-the-LSA64-Dataset.pdf"
- Description: Conversing with hearing disabled people is a significant challenge; however, computer vision advancements have significantly improved this through automated sign language recognition. One of the common issues in sign language recognition is signer-dependence, where variations arise from varying signers, who gesticulate naturally. Utilising the LSA64 dataset, a small scale Argentinian isolated sign language recognition, we investigate signer-independent sign language recognition. An InceptionV3-GRU architecture is employed to extract and classify spatial and temporal information for automated sign language recognition. The signer-dependent approach yielded an accuracy of 97.03%, whereas the signer-independent approach achieved an accuracy of 74.22%. The signer-independent system shows promise towards addressing the real-world and common issue of signer-dependence in sign language recognition.
- Full Text:
- Date Issued: 2022
Early dehydration detection using infrared imaging
- Authors: Poole, Louise C , Brown, Dane L , Connan, James
- Date: 2021
- Subjects: To be catalogued
- Language: English
- Type: text , article
- Identifier: http://hdl.handle.net/10962/465656 , vital:76629 , xlink:href="https://www.researchgate.net/profile/Louise-Poole-3/publication/357578445_Early_Dehydration_Detection_Using_Infrared_Imaging/links/61d5664eb8305f7c4b231d50/Early-Dehydration-Detection-Using-Infrared-Imaging.pdf"
- Description: Crop loss and failure have devastating impacts on a country’s economy and food security. Developing effective and inexpensive systems to minimize crop loss has become essential. Recently, multispectral imaging—in particular visible and infrared imaging—have become popular for analyzing plants and show potential for early identification of plant stress. We created a directly comparable visible and infrared image dataset for dehydration in spinach leaves. We created and compared various models trained on both datasets and concluded that the models trained on the infrared dataset outperformed all of those trained on the visible dataset. In particular, the models trained to identify early signs of dehydration yielded 45% difference in accuracy, with the infrared model obtaining 70% accuracy and the visible model obtaining 25% accuracy. Infrared imaging thus shows promising potential for application in early plant stress and disease identification.
- Full Text:
- Date Issued: 2021
An Evaluation of Text Mining Techniques in Sampling of Network Ports from IBR Traffic
- Authors: Chindipha, Stones D , Irwin, Barry V W , Herbert, Alan
- Date: 2019
- Subjects: To be catalogued
- Language: English
- Type: text , article
- Identifier: http://hdl.handle.net/10962/473740 , vital:77677 , xlink:href="https://www.researchgate.net/profile/Stones-Chindi-pha/publication/335910179_An_Evaluation_of_Text_Mining_Techniques_in_Sampling_of_Network_Ports_from_IBR_Traffic/links/5d833084458515cbd1985a38/An-Evaluation-of-Text-Mining-Techniques-in-Sampling-of-Network-Ports-from-IBR-Traffic.pdf"
- Description: Information retrieval (IR) has had techniques that have been used to gauge the extent to which certain keywords can be retrieved from a document. These techniques have been used to measure similarities in duplicated images, native language identification, optimize algorithms, among others. With this notion, this study proposes the use of four of the Information Retrieval Techniques (IRT/IR) to gauge the implications of sampling a/24 IPv4 ports into smaller subnet equivalents. Using IR, this paper shows how the ports found in a/24 IPv4 net-block relate to those found in the smaller subnet equivalents. Using Internet Background Radiation (IBR) data that was collected from Rhodes University, the study found compelling evidence of the viability of using such techniques in sampling datasets. Essentially, being able to identify the variation that comes with sampling the baseline dataset. It shows how the various samples are similar to the baseline dataset. The correlation observed in the scores proves how viable these techniques are to quantifying variations in the sampling of IBR data. In this way, one can identify which subnet equivalent best represents the unique ports found in the baseline dataset (IPv4 net-block dataset).
- Full Text:
- Date Issued: 2019
Plant disease detection and classification for farmers and everyday gardeners
- Authors: Poole, Louise C , Brown, Dane L
- Date: 2019
- Subjects: To be catalogued
- Language: English
- Type: text , article
- Identifier: http://hdl.handle.net/10962/465722 , vital:76635 , xlink:href="https://www.researchgate.net/profile/Dane-Brown-2/publication/335378684_Plant_Disease_Detection_and_Classification_for_Farmers_and_Everyday_Gardeners/links/5d611905299bf1f70b090b54/Plant-Disease-Detection-and-Classification-for-Farmers-and-Everyday-Gardeners.pdf"
- Description: Identifying and rating diseases by hand is an expensive, time consuming, subjective and unreliable method as compared to what computers can provide. Image processing and machine learning enable automated disease identification. Research has proven that automated disease identification systems can be used as a preventative measure against plant rot and death. This paper narrows down the best techniques to segment images of leaves toward improved classification of diseases found on those leaves. An investigation is conducted on image segmentation and machine learning techniques, including state-of-the-art systems, to determine the most appropriate approach to prevent death and rot in plants. Promising results were observed during testing, and show that a system can be implemented to assist with plant health that is relevant to both home gardeners and farmers.
- Full Text:
- Date Issued: 2019
Poacher detection and wildlife counting system
- Authors: Brown, Dane L , Schormann, Daniel
- Date: 2019
- Subjects: To be catalogued
- Language: English
- Type: text , article
- Identifier: http://hdl.handle.net/10962/465733 , vital:76636 , xlink:href="https://www.researchgate.net/profile/Dane-Brown-2/publication/335378767_Poacher_Detection_and_Wildlife_Counting_System/links/5d6117c7a6fdccc32ccd2cac/Poacher-Detection-and-Wildlife-Counting-System.pdf"
- Description: The illegal hunting of wildlife is a serious problem, causing a large number of animals to approach extinction or worse. Drones provide a viable option for constant surveillance and multiple instances of using drones for this purpose have been tried. However, existing methods predominantly rely on manual surveillance from camera feeds. This paper shows that using either visible or thermal cameras, with modern image processing and machine learning techniques, enables a system to autonomously detect humans, while tracking animals by their identity number (id). The thermal characteristics of special but inexpensive cameras are used for object detection with centroid tracking, and convolutional neural networks are used to classify humans and wildlife. Classification also enables the counting of wildlife by id, which can help game reserves keep track of wildlife.
- Full Text:
- Date Issued: 2019
Virtual Gym Instructor
- Authors: Brown, Dane L , Ndleve, Mixo
- Date: 2019
- Subjects: To be catalogued
- Language: English
- Type: text , article
- Identifier: http://hdl.handle.net/10962/465744 , vital:76637 , xlink:href="https://www.researchgate.net/profile/Dane-Brown-2/publication/335378603_Virtual_Gym_Instructor/links/5d6118a892851c619d7268c1/Virtual-Gym-Instructor.pdf"
- Description: The fourth industrial revolution and the continuous development of new technologies have presented a golden platter for sedentary living. Noncommunicable diseases such as, cancers, cardiovascular and respiratory deficiencies, and diabetes have reached epidemic levels as a consequence. A traditional gym instructor screens clients to prescribe exercise programs that can help them lower the risk of noncommunicable lifestyle diseases. However, gym instructors often come at a cost and are not always affordable, available or accessible. This research investigated whether modern computing power can be utilized to develop a system in the form of a cost effective alternative exercise program – Virtual Gym Instructor. The system demonstrated perfect realtime object detection and tracking up to four metres away from the camera and produced results for distances up to eight metres away.
- Full Text:
- Date Issued: 2019
Effectiveness of Sampling a Small Sized Network Telescope in Internet Background Radiation Data Collection
- Authors: Chindipha, Stones D , Irwin, Barry V W , Herbert, Alan
- Date: 2018
- Subjects: To be catalogued
- Language: English
- Type: text , article
- Identifier: http://hdl.handle.net/10962/473783 , vital:77681 , xlink:href="https://www.researchgate.net/profile/Barry-Irwin/publication/327624431_Effectiveness_of_Sampling_a_Small_Sized_Network_Telescope_in_Internet_Background_Radiation_Data_Collection/links/5b9a5067299bf14ad4d793a1/Effectiveness-of-Sampling-a-Small-Sized-Network-Telescope-in-Internet-Background-Radiation-Data-Collection.pdf"
- Description: What is known today as the modern Internet has long relied on the existence of, and use of, IPv4 addresses. However, due to the rapid growth of the Internet of Things (IoT), and limited address space within IPv4, acquiring large IPv4 subnetworks is becoming increasingly difficult. The exhaustion of the IPv4 address space has made it near impossible for organizations to gain access to large blocks of IP space. This is of great concern particularly in the security space which often relies on acquiring large network blocks for performing a technique called Internet Background Radiation (IBR) monitoring. This technique monitors IPv4 addresses which have no services running on them. In practice, no traffic should ever arrive at such an IPv4 address, and so is marked as an anomaly, and thus recorded and analyzed. This research aims to address the problem brought forth by IPv4 address space exhaustion in relation to IBR monitoring. This study’s intent is to identify the smallest subnet that best represents the attributes found in the/24 IPv4 address. This is done by determining how well a subset of the monitored original subnetwork represents the information gathered by the original subnetwork. Determining the best method of selecting a subset of IPv4 addresses from a subnetwork will enable IBR research to continue in the best way possible in an ever restricting research space.
- Full Text:
- Date Issued: 2018
Efficient Biometric Access Control for Larger Scale Populations
- Authors: Brown, Dane L , Bradshaw, Karen L
- Date: 2018
- Subjects: To be catalogued
- Language: English
- Type: text , article
- Identifier: http://hdl.handle.net/10962/465667 , vital:76630 , xlink:href="https://www.researchgate.net/profile/Dane-Brown-2/publication/335378829_Efficient_Biometric_Access_Control_for_Larger_Scale_Populations/links/5d61159ea6fdccc32ccd2c8a/Efficient-Biometric-Access-Control-for-Larger-Scale-Populations.pdf"
- Description: Biometric applications and databases are growing at an alarming rate. Processing large or complex biometric data induces longer wait times that can limit usability during application. This paper focuses on increasing the processing speed of biometric data, and calls for a parallel approach to data processing that is beyond the capability of a central processing unit (CPU). The graphical processing unit (GPU) is effectively utilized with compute unified device architecture (CUDA), and results in at least triple the processing speed when compared with a previously presented accurate and secure multimodal biometric system. When saturating the CPU-only implementation with more individuals than the available thread count, the GPU-assisted implementation outperforms it exponentially. The GPU-assisted implementation is also validated to have the same accuracy of the original system, and thus shows promising advancements in both accuracy and processing speed in the challenging big data world.
- Full Text:
- Date Issued: 2018
An analysis on the re-emergence of SQL Slammer worm using network telescope data
- Authors: Chindipha, Stones D , Irwin, Barry V W
- Date: 2017
- Subjects: To be catalogued
- Language: English
- Type: text , article
- Identifier: http://hdl.handle.net/10962/473718 , vital:77675 , xlink:href="https://www.researchgate.net/publication/327622806_An_Analysis_on_the_Re-emergence_of_SQL_Slammer_Worm_Using_Network_Telescope_Data"
- Description: The SQL Slammer worm is a self propagated computer virus that caused a denial of service on some Internet hosts and dramatically slowed down general Internet traffic. An observation of network traffic captured in the Rhodes University’s network telescopes shows that traffic observed in it shows an escalation in the number of packets captured by the telescopes between January 2014 and December 2016 when the expected traffic was meant to take a constant decline in UDP packets from port 1434. Using data captured over a period of 84 months, the analysis done in this study identified top ten /24 source IP addresses that Slammer worm repeatedly used for this attack together with their geolocation. It also shows the trend of UDP 1434 packets received by the two network telescopes from January 2009 to December 2015. In line with epidemic model, the paper has shown how this traffic fits in as SQL Slammer worm attack. Consistent number of packets observed in the two telescopes between 2014 and 2016 shows qualities of the Slammer worm attack. Basic time series and decomposition of additive time series graphs have been used to show trend and observed UDP packets over the time frame of study.
- Full Text:
- Date Issued: 2017
Enhanced biometric access control for mobile devices
- Authors: Brown, Dane L , Bradshaw, Karen L
- Date: 2017
- Subjects: To be catalogued
- Language: English
- Type: text , article
- Identifier: http://hdl.handle.net/10962/465678 , vital:76631
- Description: In the new Digital Economy, mobile devices are increasingly 978-0-620-76756-9being used for tasks that involve sensitive and/or financial data. Hitherto, security on smartphones has not been a priority and furthermore, users tend to ignore the security features in favour of more rapid access to the device. We propose an authentication system that can provide enhanced security by utilizing multi-modal biometrics from a single image, captured at arm’s length, containing unique face and iris data. The system is compared to state-of-the-art face and iris recognition systems, in related studies using the CASIA-Iris-Distance dataset and the IITD iris dataset. The proposed system outperforms the related studies in all experiments and shows promising advancements to at-a-distance iris recognition on mobile devices.
- Full Text:
- Date Issued: 2017
A dynamically weighted multi-modal biometric security system
- Authors: Brown, Dane L , Bradshaw, Karen L
- Date: 2016
- Subjects: To be catalogued
- Language: English
- Type: text , article
- Identifier: http://hdl.handle.net/10962/473684 , vital:77672 , xlink:href="https://www.researchgate.net/publication/315839228_A_Dynamically_Weighted_Multi-Modal_Biometric_Security_System"
- Description: The face, fingerprint and palmprint feature vectors are automatically extracted and dynamically selected for fusion at the feature-level, toward an improved human identification accuracy. The feature-level has a higher potential accuracy than the match score-level. However, leveraging this potential requires a new approach. This work demonstrates a novel dynamic weighting algorithm for improved image-based biometric feature-fusion. A comparison is performed on uni-modal, bi-modal, tri-modal and proposed dynamic approaches. The proposed dynamic approach yields a high genuine acceptance rate of 99.25% genuine acceptance rate at a false acceptance rate of 1% on challenging datasets and big impostor datasets.
- Full Text:
- Date Issued: 2016