A Practical Use for AI-Generated Images
- Boby, Alden, Brown, Dane L, Connan, James
- Authors: Boby, Alden , Brown, Dane L , Connan, James
- Date: 2023
- Subjects: To be catalogued
- Language: English
- Type: text , article
- Identifier: http://hdl.handle.net/10962/463345 , vital:76401 , xlink:href="https://link.springer.com/chapter/10.1007/978-3-031-43838-7_12"
- Description: Collecting data for research can be costly and time-consuming, and available methods to speed up the process are limited. This research paper compares real data and AI-generated images for training an object detection model. The study aimed to assess how the utilisation of AI-generated images influences the performance of an object detection model. The study used a popular object detection model, YOLO, and trained it on a dataset with real car images as well as a synthetic dataset generated with a state-of-the-art diffusion model. The results showed that while the model trained on real data performed better on real-world images, the model trained on AI-generated images, in some cases, showed improved performance on certain images and was good enough to function as a licence plate detector on its own. The study highlights the potential of using AI-generated images for data augmentation in object detection models and sheds light on the trade-off between real and synthetic data in the training process. The findings of this study can inform future research in object detection and help practitioners make informed decisions when choosing between real and synthetic data for training object detection models.
- Full Text:
- Date Issued: 2023
- Authors: Boby, Alden , Brown, Dane L , Connan, James
- Date: 2023
- Subjects: To be catalogued
- Language: English
- Type: text , article
- Identifier: http://hdl.handle.net/10962/463345 , vital:76401 , xlink:href="https://link.springer.com/chapter/10.1007/978-3-031-43838-7_12"
- Description: Collecting data for research can be costly and time-consuming, and available methods to speed up the process are limited. This research paper compares real data and AI-generated images for training an object detection model. The study aimed to assess how the utilisation of AI-generated images influences the performance of an object detection model. The study used a popular object detection model, YOLO, and trained it on a dataset with real car images as well as a synthetic dataset generated with a state-of-the-art diffusion model. The results showed that while the model trained on real data performed better on real-world images, the model trained on AI-generated images, in some cases, showed improved performance on certain images and was good enough to function as a licence plate detector on its own. The study highlights the potential of using AI-generated images for data augmentation in object detection models and sheds light on the trade-off between real and synthetic data in the training process. The findings of this study can inform future research in object detection and help practitioners make informed decisions when choosing between real and synthetic data for training object detection models.
- Full Text:
- Date Issued: 2023
An Evaluation of Machine Learning Methods for Classifying Bot Traffic in Software Defined Networks
- Van Staden, Joshua, Brown, Dane L
- Authors: Van Staden, Joshua , Brown, Dane L
- Date: 2023
- Subjects: To be catalogued
- Language: English
- Type: text , article
- Identifier: http://hdl.handle.net/10962/463357 , vital:76402 , xlink:href="https://link.springer.com/chapter/10.1007/978-981-19-7874-6_72"
- Description: Internet security is an ever-expanding field. Cyber-attacks occur very frequently, and so detecting them is an important aspect of preserving services. Machine learning offers a helpful tool with which to detect cyber attacks. However, it is impossible to deploy a machine-learning algorithm to detect attacks in a non-centralized network. Software Defined Networks (SDNs) offer a centralized view of a network, allowing machine learning algorithms to detect malicious activity within a network. The InSDN dataset is a recently-released dataset that contains a set of sniffed packets within a virtual SDN. These sniffed packets correspond to various attacks, including DDoS attacks, Probing and Password-Guessing, among others. This study aims to evaluate various machine learning models against this new dataset. Specifically, we aim to evaluate their classification ability and runtimes when trained on fewer features. The machine learning models tested include a Neural Network, Support Vector Machine, Random Forest, Multilayer Perceptron, Logistic Regression, and K-Nearest Neighbours. Cluster-based algorithms such as the K-Nearest Neighbour and Random Forest proved to be the best performers. Linear-based algorithms such as the Multilayer Perceptron performed the worst. This suggests a good level of clustering in the top few features with little space for linear separability. The reduction of features significantly reduced training time, particularly in the better-performing models.
- Full Text:
- Date Issued: 2023
- Authors: Van Staden, Joshua , Brown, Dane L
- Date: 2023
- Subjects: To be catalogued
- Language: English
- Type: text , article
- Identifier: http://hdl.handle.net/10962/463357 , vital:76402 , xlink:href="https://link.springer.com/chapter/10.1007/978-981-19-7874-6_72"
- Description: Internet security is an ever-expanding field. Cyber-attacks occur very frequently, and so detecting them is an important aspect of preserving services. Machine learning offers a helpful tool with which to detect cyber attacks. However, it is impossible to deploy a machine-learning algorithm to detect attacks in a non-centralized network. Software Defined Networks (SDNs) offer a centralized view of a network, allowing machine learning algorithms to detect malicious activity within a network. The InSDN dataset is a recently-released dataset that contains a set of sniffed packets within a virtual SDN. These sniffed packets correspond to various attacks, including DDoS attacks, Probing and Password-Guessing, among others. This study aims to evaluate various machine learning models against this new dataset. Specifically, we aim to evaluate their classification ability and runtimes when trained on fewer features. The machine learning models tested include a Neural Network, Support Vector Machine, Random Forest, Multilayer Perceptron, Logistic Regression, and K-Nearest Neighbours. Cluster-based algorithms such as the K-Nearest Neighbour and Random Forest proved to be the best performers. Linear-based algorithms such as the Multilayer Perceptron performed the worst. This suggests a good level of clustering in the top few features with little space for linear separability. The reduction of features significantly reduced training time, particularly in the better-performing models.
- Full Text:
- Date Issued: 2023
Darknet Traffic Detection Using Histogram-Based Gradient Boosting
- Brown, Dane L, Sepula, Chikondi
- Authors: Brown, Dane L , Sepula, Chikondi
- Date: 2023
- Subjects: To be catalogued
- Language: English
- Type: text , article
- Identifier: http://hdl.handle.net/10962/464063 , vital:76472 , xlink:href="https://link.springer.com/chapter/10.1007/978-981-99-1624-5_59"
- Description: The network security sector has observed a rise in severe attacks emanating from the darknet or encrypted networks in recent years. Network intrusion detection systems (NIDS) capable of detecting darknet or encrypted traffic must be developed to increase system security. Machine learning algorithms can effectively detect darknet activities when trained on encrypted and conventional network data. However, the performance of the system may be influenced, among other things, by the choice of machine learning models, data preparation techniques, and feature selection methodologies. The histogram-based gradient boosting strategy known as categorical boosting (CatBoost) was tested to see how well it could find darknet traffic. The performance of the model was examined using feature selection strategies such as correlation coefficient, variance threshold, SelectKBest, and recursive feature removal (RFE). Following the categorization of traffic as “darknet” or “regular”, a multi-class classification was used to determine the software application associated with the traffic. Further study was carried out on well-known machine learning methods such as random forests (RF), decision trees (DT), linear support vector classifier (SVC Linear), and long-short term memory (LST) (LSTM). The proposed model achieved good results with 98.51% binary classification accuracy and 88% multi-class classification accuracy.
- Full Text:
- Date Issued: 2023
- Authors: Brown, Dane L , Sepula, Chikondi
- Date: 2023
- Subjects: To be catalogued
- Language: English
- Type: text , article
- Identifier: http://hdl.handle.net/10962/464063 , vital:76472 , xlink:href="https://link.springer.com/chapter/10.1007/978-981-99-1624-5_59"
- Description: The network security sector has observed a rise in severe attacks emanating from the darknet or encrypted networks in recent years. Network intrusion detection systems (NIDS) capable of detecting darknet or encrypted traffic must be developed to increase system security. Machine learning algorithms can effectively detect darknet activities when trained on encrypted and conventional network data. However, the performance of the system may be influenced, among other things, by the choice of machine learning models, data preparation techniques, and feature selection methodologies. The histogram-based gradient boosting strategy known as categorical boosting (CatBoost) was tested to see how well it could find darknet traffic. The performance of the model was examined using feature selection strategies such as correlation coefficient, variance threshold, SelectKBest, and recursive feature removal (RFE). Following the categorization of traffic as “darknet” or “regular”, a multi-class classification was used to determine the software application associated with the traffic. Further study was carried out on well-known machine learning methods such as random forests (RF), decision trees (DT), linear support vector classifier (SVC Linear), and long-short term memory (LST) (LSTM). The proposed model achieved good results with 98.51% binary classification accuracy and 88% multi-class classification accuracy.
- Full Text:
- Date Issued: 2023
Early Plant Disease Detection using Infrared and Mobile Photographs in Natural Environment
- De Silva, Malitha, Brown, Dane L
- Authors: De Silva, Malitha , Brown, Dane L
- Date: 2023
- Subjects: To be catalogued
- Language: English
- Type: text , article
- Identifier: http://hdl.handle.net/10962/464085 , vital:76474 , xlink:href="https://link.springer.com/chapter/10.1007/978-3-031-37717-4_21"
- Description: Plant disease identification is a critical aspect of plant health management. Identifying plant diseases is challenging since they manifest themselves in various forms and tend to occur when the plant is still in its juvenile stage. Plant disease also has cascading effects on food security, livelihoods and the environment’s safety, so early detection is vital. This work demonstrates the effectiveness of mobile and multispectral images captured in viable and Near Infrared (NIR) ranges to identify plant diseases under realistic environmental conditions. The data sets were classified using popular CNN models Xception, DenseNet121 and ResNet50V2, resulting in greater than 92% training and 74% test accuracy for all the data collected using various Kolari vision lenses. Moreover, an openly available balanced data set was used to compare the effect of the data set balance and unbalanced characteristics on the classification accuracy. The result showed that balanced data sets do not impact the outcome.
- Full Text:
- Date Issued: 2023
- Authors: De Silva, Malitha , Brown, Dane L
- Date: 2023
- Subjects: To be catalogued
- Language: English
- Type: text , article
- Identifier: http://hdl.handle.net/10962/464085 , vital:76474 , xlink:href="https://link.springer.com/chapter/10.1007/978-3-031-37717-4_21"
- Description: Plant disease identification is a critical aspect of plant health management. Identifying plant diseases is challenging since they manifest themselves in various forms and tend to occur when the plant is still in its juvenile stage. Plant disease also has cascading effects on food security, livelihoods and the environment’s safety, so early detection is vital. This work demonstrates the effectiveness of mobile and multispectral images captured in viable and Near Infrared (NIR) ranges to identify plant diseases under realistic environmental conditions. The data sets were classified using popular CNN models Xception, DenseNet121 and ResNet50V2, resulting in greater than 92% training and 74% test accuracy for all the data collected using various Kolari vision lenses. Moreover, an openly available balanced data set was used to compare the effect of the data set balance and unbalanced characteristics on the classification accuracy. The result showed that balanced data sets do not impact the outcome.
- Full Text:
- Date Issued: 2023
Efficient Plant Disease Detection and Classification for Android
- Brown, Dane L, Mazibuko, Sifisokuhle
- Authors: Brown, Dane L , Mazibuko, Sifisokuhle
- Date: 2023
- Subjects: To be catalogued
- Language: English
- Type: text , article
- Identifier: http://hdl.handle.net/10962/464096 , vital:76475 , xlink:href="https://link.springer.com/chapter/10.1007/978-981-99-1624-5_39"
- Description: This paper investigates the feasibility of using a CNN model to diagnose plant diseases in the wild. Plant diseases are a major risk to ecosystems, human and animal health, and the quality of life overall. They may reduce farm productivity drastically, leaving farmers with financial losses and food insecurity. Small-scale farmers and producers cannot pay for an expert to look at their plants for plant diseases because it would cost too much. A mobile solution is thus built for the Android platform that utilises a unified deep learning model to diagnose plant diseases and provide farmers with treatment information. The literature-recommended CNN architectures were first analysed on the PlantVillage dataset, and the best-performing model was trained for integration into the application. While training on the tomato subset of the PlantVillage dataset, the VGG16 and InceptionV3 networks achieved a higher F1-score of 94.49% than the MobileNetsV3Large and EfficientNetB0 networks (without parameter tuning). The VGG model achieved 94.43% accuracy and 0.24 loss on the RGB PlantVillage dataset, outperforming the segmented and greyscaled datasets, and was therefore chosen for use in the application. When tested on complex data collected in the wild, the VGG16 model trained on the RGB dataset yielded an accuracy of 63.02%. Thus, this research revealed the discrepancy between simple and real-world data, as well as the viability of present methodologies for future research.
- Full Text:
- Date Issued: 2023
- Authors: Brown, Dane L , Mazibuko, Sifisokuhle
- Date: 2023
- Subjects: To be catalogued
- Language: English
- Type: text , article
- Identifier: http://hdl.handle.net/10962/464096 , vital:76475 , xlink:href="https://link.springer.com/chapter/10.1007/978-981-99-1624-5_39"
- Description: This paper investigates the feasibility of using a CNN model to diagnose plant diseases in the wild. Plant diseases are a major risk to ecosystems, human and animal health, and the quality of life overall. They may reduce farm productivity drastically, leaving farmers with financial losses and food insecurity. Small-scale farmers and producers cannot pay for an expert to look at their plants for plant diseases because it would cost too much. A mobile solution is thus built for the Android platform that utilises a unified deep learning model to diagnose plant diseases and provide farmers with treatment information. The literature-recommended CNN architectures were first analysed on the PlantVillage dataset, and the best-performing model was trained for integration into the application. While training on the tomato subset of the PlantVillage dataset, the VGG16 and InceptionV3 networks achieved a higher F1-score of 94.49% than the MobileNetsV3Large and EfficientNetB0 networks (without parameter tuning). The VGG model achieved 94.43% accuracy and 0.24 loss on the RGB PlantVillage dataset, outperforming the segmented and greyscaled datasets, and was therefore chosen for use in the application. When tested on complex data collected in the wild, the VGG16 model trained on the RGB dataset yielded an accuracy of 63.02%. Thus, this research revealed the discrepancy between simple and real-world data, as well as the viability of present methodologies for future research.
- Full Text:
- Date Issued: 2023
Enabling Vehicle Search Through Robust Licence Plate Detection
- Boby, Alden, Brown, Dane L, Connan, James, Marais, Marc, Kuhlane, Luxolo L
- Authors: Boby, Alden , Brown, Dane L , Connan, James , Marais, Marc , Kuhlane, Luxolo L
- Date: 2023
- Subjects: To be catalogued
- Language: English
- Type: text , article
- Identifier: http://hdl.handle.net/10962/463372 , vital:76403 , xlink:href="https://ieeexplore.ieee.org/abstract/document/10220508"
- Description: Licence plate recognition has many practical applications for security and surveillance. This paper presents a robust licence plate detection system that uses string-matching algorithms to identify a vehicle in data. Object detection models have had limited application in the character recognition domain. The system utilises the YOLO object detection model to perform character recognition to ensure more accurate character predictions. The model incorporates super-resolution techniques to enhance the quality of licence plate images to increase character recognition accuracy. The proposed system can accurately detect license plates in diverse conditions and can handle license plates with varying fonts and backgrounds. The system's effectiveness is demonstrated through experimentation on components of the system, showing promising license plate detection and character recognition accuracy. The overall system works with all the components to track vehicles by matching a target string with detected licence plates in a scene. The system has potential applications in law enforcement, traffic management, and parking systems and can significantly advance surveillance and security through automation.
- Full Text:
- Date Issued: 2023
- Authors: Boby, Alden , Brown, Dane L , Connan, James , Marais, Marc , Kuhlane, Luxolo L
- Date: 2023
- Subjects: To be catalogued
- Language: English
- Type: text , article
- Identifier: http://hdl.handle.net/10962/463372 , vital:76403 , xlink:href="https://ieeexplore.ieee.org/abstract/document/10220508"
- Description: Licence plate recognition has many practical applications for security and surveillance. This paper presents a robust licence plate detection system that uses string-matching algorithms to identify a vehicle in data. Object detection models have had limited application in the character recognition domain. The system utilises the YOLO object detection model to perform character recognition to ensure more accurate character predictions. The model incorporates super-resolution techniques to enhance the quality of licence plate images to increase character recognition accuracy. The proposed system can accurately detect license plates in diverse conditions and can handle license plates with varying fonts and backgrounds. The system's effectiveness is demonstrated through experimentation on components of the system, showing promising license plate detection and character recognition accuracy. The overall system works with all the components to track vehicles by matching a target string with detected licence plates in a scene. The system has potential applications in law enforcement, traffic management, and parking systems and can significantly advance surveillance and security through automation.
- Full Text:
- Date Issued: 2023
Enhanced plant species and early water stress detection using visible and near-infrared spectra
- Brown, Dane L, Poole, Louise C
- Authors: Brown, Dane L , Poole, Louise C
- Date: 2023
- Subjects: To be catalogued
- Language: English
- Type: text , article
- Identifier: http://hdl.handle.net/10962/463384 , vital:76404 , xlink:href="https://link.springer.com/chapter/10.1007/978-981-19-9819-5_55"
- Description: This paper reports on recent successful work aimed at preventing crop loss and failure before visible symptoms are present. Food security is critical, especially after the COVID-19 pandemic. Detecting early-stage plant stresses in agriculture is essential in minimizing crop damage and maximizing yield. Identification of both the stress type and cause is a non-trivial multitask classification problem. However, the application of spectroscopy to early plant diseases and stress detection has become viable with recent advancements in technology. Suitable frequencies of the electromagnetic spectrum and machine learning algorithms were thus first investigated. This guided data collection in two sessions by capturing standard visible images in contrast with images from multiple spectra (VIS-IR). These images consisted of six plant species that were carefully monitored from healthy to dehydrated stages. Promising results were achieved using VIS-IR compared to standard visible images on three deep learning architectures. Statistically, significant accuracy improvements were shown for VIS-IR for early dehydration detection, where ResNet-44 modelling of VIS-IR input yielded 92.5% accuracy compared to 77.5% on visible input on general plant species. Moreover, ResNet-44 achieved good species separation.
- Full Text:
- Date Issued: 2023
- Authors: Brown, Dane L , Poole, Louise C
- Date: 2023
- Subjects: To be catalogued
- Language: English
- Type: text , article
- Identifier: http://hdl.handle.net/10962/463384 , vital:76404 , xlink:href="https://link.springer.com/chapter/10.1007/978-981-19-9819-5_55"
- Description: This paper reports on recent successful work aimed at preventing crop loss and failure before visible symptoms are present. Food security is critical, especially after the COVID-19 pandemic. Detecting early-stage plant stresses in agriculture is essential in minimizing crop damage and maximizing yield. Identification of both the stress type and cause is a non-trivial multitask classification problem. However, the application of spectroscopy to early plant diseases and stress detection has become viable with recent advancements in technology. Suitable frequencies of the electromagnetic spectrum and machine learning algorithms were thus first investigated. This guided data collection in two sessions by capturing standard visible images in contrast with images from multiple spectra (VIS-IR). These images consisted of six plant species that were carefully monitored from healthy to dehydrated stages. Promising results were achieved using VIS-IR compared to standard visible images on three deep learning architectures. Statistically, significant accuracy improvements were shown for VIS-IR for early dehydration detection, where ResNet-44 modelling of VIS-IR input yielded 92.5% accuracy compared to 77.5% on visible input on general plant species. Moreover, ResNet-44 achieved good species separation.
- Full Text:
- Date Issued: 2023
Exploring the Incremental Improvements of YOLOv5 on Tracking and Identifying Great White Sharks in Cape Town
- Kuhlane, Luxolo L, Brown, Dane L, Boby, Alden
- Authors: Kuhlane, Luxolo L , Brown, Dane L , Boby, Alden
- Date: 2023
- Subjects: To be catalogued
- Language: English
- Type: text , article
- Identifier: http://hdl.handle.net/10962/464107 , vital:76476 , xlink:href="https://link.springer.com/chapter/10.1007/978-3-031-37963-5_98"
- Description: The information on great white sharks is used by scientists to help better understand the marine organisms and to mitigate any chances of extinction of great white sharks. Sharks play a very important role in the ocean, and their role in the oceans is under-appreciated by the general public, which results in negative attitudes towards sharks. The tracking and identification of sharks are done using manual labour, which is not very accurate and time-consuming. This paper uses a deep learning approach to help identify and track great white sharks in Cape Town. A popular object detecting system used in this paper is YOLO, which is implemented to help identify the great white shark. In conjunction with YOLO, the paper also uses ESRGAN to help upscale low-quality images from the datasets into more high-quality images before being put into the YOLO system. The main focus of this paper is to help train the system; this includes training the system to identify great white sharks in difficult conditions such as murky water or unclear deep-sea conditions.
- Full Text:
- Date Issued: 2023
- Authors: Kuhlane, Luxolo L , Brown, Dane L , Boby, Alden
- Date: 2023
- Subjects: To be catalogued
- Language: English
- Type: text , article
- Identifier: http://hdl.handle.net/10962/464107 , vital:76476 , xlink:href="https://link.springer.com/chapter/10.1007/978-3-031-37963-5_98"
- Description: The information on great white sharks is used by scientists to help better understand the marine organisms and to mitigate any chances of extinction of great white sharks. Sharks play a very important role in the ocean, and their role in the oceans is under-appreciated by the general public, which results in negative attitudes towards sharks. The tracking and identification of sharks are done using manual labour, which is not very accurate and time-consuming. This paper uses a deep learning approach to help identify and track great white sharks in Cape Town. A popular object detecting system used in this paper is YOLO, which is implemented to help identify the great white shark. In conjunction with YOLO, the paper also uses ESRGAN to help upscale low-quality images from the datasets into more high-quality images before being put into the YOLO system. The main focus of this paper is to help train the system; this includes training the system to identify great white sharks in difficult conditions such as murky water or unclear deep-sea conditions.
- Full Text:
- Date Issued: 2023
Exploring The Incremental Improvements of YOLOv7 on Bull Sharks in Mozambique
- Kuhlane, Luxolo L, Brown, Dane L, Brown, Alden
- Authors: Kuhlane, Luxolo L , Brown, Dane L , Brown, Alden
- Date: 2023
- Subjects: To be catalogued
- Language: English
- Type: text , article
- Identifier: http://hdl.handle.net/10962/464118 , vital:76478 , xlink:href="https://www.researchgate.net/profile/Dane-Brown-2/publication/368455814_Exploring_The_Incremental_Improvements_of_YOLOv7_on_Bull_Sharks_in_Mozambique/links/63e8d321dea6121757a4ba7f/Exploring-The-Incremental-Improvements-of-YOLOv7-on-Bull-Sharks-in-Mozambique.pdf?origin=journalDetailand_tp=eyJwYWdlIjoiam91cm5hbERldGFpbCJ9"
- Description: Scientists use bull shark data to better understand marine organisms and to reduce the likelihood of bull shark extinction. Sharks play an important role in the ocean, and their importance is underappreciated by the general public, leading to negative attitudes toward sharks. The tracking and identification of sharks is done by hand, which is inefficient and time-consuming. This paper employs a deep learning approach to assist in the identification and tracking of bull sharks in Mozambique. YOLO is a popular object detection system used in this paper to aid in the identification of the great white shark. In addition to YOLO, the paper employs ESRGAN to help upscale low-quality images from the datasets into higher-quality images before they are fed into the YOLO system. The primary goal of this paper is to assist in training the system to identify bull sharks in difficult conditions such as murky water or unclear deep-sea conditions.
- Full Text:
- Date Issued: 2023
- Authors: Kuhlane, Luxolo L , Brown, Dane L , Brown, Alden
- Date: 2023
- Subjects: To be catalogued
- Language: English
- Type: text , article
- Identifier: http://hdl.handle.net/10962/464118 , vital:76478 , xlink:href="https://www.researchgate.net/profile/Dane-Brown-2/publication/368455814_Exploring_The_Incremental_Improvements_of_YOLOv7_on_Bull_Sharks_in_Mozambique/links/63e8d321dea6121757a4ba7f/Exploring-The-Incremental-Improvements-of-YOLOv7-on-Bull-Sharks-in-Mozambique.pdf?origin=journalDetailand_tp=eyJwYWdlIjoiam91cm5hbERldGFpbCJ9"
- Description: Scientists use bull shark data to better understand marine organisms and to reduce the likelihood of bull shark extinction. Sharks play an important role in the ocean, and their importance is underappreciated by the general public, leading to negative attitudes toward sharks. The tracking and identification of sharks is done by hand, which is inefficient and time-consuming. This paper employs a deep learning approach to assist in the identification and tracking of bull sharks in Mozambique. YOLO is a popular object detection system used in this paper to aid in the identification of the great white shark. In addition to YOLO, the paper employs ESRGAN to help upscale low-quality images from the datasets into higher-quality images before they are fed into the YOLO system. The primary goal of this paper is to assist in training the system to identify bull sharks in difficult conditions such as murky water or unclear deep-sea conditions.
- Full Text:
- Date Issued: 2023
Learning Movement Patterns for Improving the Skills of Beginner Level Players in Competitive MOBAs
- Brown, Dane L, Bischof, Jonah
- Authors: Brown, Dane L , Bischof, Jonah
- Date: 2023
- Subjects: To be catalogued
- Language: English
- Type: text , article
- Identifier: http://hdl.handle.net/10962/464161 , vital:76482 , xlink:href="https://link.springer.com/chapter/10.1007/978-981-99-1624-5_45"
- Description: League of Legends is a massively multiplayer online battle arena (MOBA)—a form of online competitive game in which teams of five players battle to demolish the opponent’s base. Expert players are aware of when to target, how to maximise their gold, and how to make choices. These are some of the talents that distinguish them from novices. The Riot API enables the retrieval of current League of Legends game data. This data is used to construct machine learning models that can benefit amateur players. Kills and goals can assist seasoned players understand how to take advantage of micro- and macro-teams. By understanding how professional players differ from novices, we may build tools to assist novices’ decision-making. 19 of 20 games for training a random forest (RF) and decision tree (DT) regressor produced encouraging results. An unseen game was utilised to evaluate the impartiality of the findings. RF and DT correctly predicted the locations of all game events in Experiment 1 with MSEs of 9.5 and 10.6. The purpose of the previous experiment was to fine-tune when novice players deviate from professional player behaviour and establish a solid commencement for battles. Based on this discrepancy, the system provided the player with reliable recommendations on which quadrant they should be in and which event/objective they should complete. This has shown to be a beneficial method for modelling player behaviour in future research.
- Full Text:
- Date Issued: 2023
- Authors: Brown, Dane L , Bischof, Jonah
- Date: 2023
- Subjects: To be catalogued
- Language: English
- Type: text , article
- Identifier: http://hdl.handle.net/10962/464161 , vital:76482 , xlink:href="https://link.springer.com/chapter/10.1007/978-981-99-1624-5_45"
- Description: League of Legends is a massively multiplayer online battle arena (MOBA)—a form of online competitive game in which teams of five players battle to demolish the opponent’s base. Expert players are aware of when to target, how to maximise their gold, and how to make choices. These are some of the talents that distinguish them from novices. The Riot API enables the retrieval of current League of Legends game data. This data is used to construct machine learning models that can benefit amateur players. Kills and goals can assist seasoned players understand how to take advantage of micro- and macro-teams. By understanding how professional players differ from novices, we may build tools to assist novices’ decision-making. 19 of 20 games for training a random forest (RF) and decision tree (DT) regressor produced encouraging results. An unseen game was utilised to evaluate the impartiality of the findings. RF and DT correctly predicted the locations of all game events in Experiment 1 with MSEs of 9.5 and 10.6. The purpose of the previous experiment was to fine-tune when novice players deviate from professional player behaviour and establish a solid commencement for battles. Based on this discrepancy, the system provided the player with reliable recommendations on which quadrant they should be in and which event/objective they should complete. This has shown to be a beneficial method for modelling player behaviour in future research.
- Full Text:
- Date Issued: 2023
Multispectral Plant Disease Detection with Vision Transformer–Convolutional Neural Network Hybrid Approaches
- De Silva, Malitha, Brown, Dane L
- Authors: De Silva, Malitha , Brown, Dane L
- Date: 2023
- Subjects: To be catalogued
- Language: English
- Type: text , article
- Identifier: http://hdl.handle.net/10962/463428 , vital:76408 , xlink:href="https://doi.org/10.3390/s23208531"
- Description: Plant diseases pose a critical threat to global agricultural productivity, demanding timely detection for effective crop yield management. Traditional methods for disease identification are laborious and require specialised expertise. Leveraging cutting-edge deep learning algorithms, this study explores innovative approaches to plant disease identification, combining Convolutional Neural Networks (CNNs) and Vision Transformers (ViTs) to enhance accuracy. A multispectral dataset was meticulously collected to facilitate this research using six 50 mm filter filters, covering both the visible and several near-infrared (NIR) wavelengths. Among the models employed, ViT-B16 notably achieved the highest test accuracy, precision, recall, and F1 score across all filters, with averages of 83.3%, 90.1%, 90.75%, and 89.5%, respectively. Furthermore, a comparative analysis highlights the pivotal role of balanced datasets in selecting the appropriate wavelength and deep learning model for robust disease identification. These findings promise to advance crop disease management in real-world agricultural applications and contribute to global food security. The study underscores the significance of machine learning in transforming plant disease diagnostics and encourages further research in this field.
- Full Text:
- Date Issued: 2023
- Authors: De Silva, Malitha , Brown, Dane L
- Date: 2023
- Subjects: To be catalogued
- Language: English
- Type: text , article
- Identifier: http://hdl.handle.net/10962/463428 , vital:76408 , xlink:href="https://doi.org/10.3390/s23208531"
- Description: Plant diseases pose a critical threat to global agricultural productivity, demanding timely detection for effective crop yield management. Traditional methods for disease identification are laborious and require specialised expertise. Leveraging cutting-edge deep learning algorithms, this study explores innovative approaches to plant disease identification, combining Convolutional Neural Networks (CNNs) and Vision Transformers (ViTs) to enhance accuracy. A multispectral dataset was meticulously collected to facilitate this research using six 50 mm filter filters, covering both the visible and several near-infrared (NIR) wavelengths. Among the models employed, ViT-B16 notably achieved the highest test accuracy, precision, recall, and F1 score across all filters, with averages of 83.3%, 90.1%, 90.75%, and 89.5%, respectively. Furthermore, a comparative analysis highlights the pivotal role of balanced datasets in selecting the appropriate wavelength and deep learning model for robust disease identification. These findings promise to advance crop disease management in real-world agricultural applications and contribute to global food security. The study underscores the significance of machine learning in transforming plant disease diagnostics and encourages further research in this field.
- Full Text:
- Date Issued: 2023
Plant Disease Detection using Vision Transformers on Multispectral Natural Environment Images
- De Silva, Malitha, Brown, Dane L
- Authors: De Silva, Malitha , Brown, Dane L
- Date: 2023
- Subjects: To be catalogued
- Language: English
- Type: text , article
- Identifier: http://hdl.handle.net/10962/463456 , vital:76410 , xlink:href="https://ieeexplore.ieee.org/abstract/document/10220517"
- Description: Enhancing agricultural practices has become essential in mitigating global hunger. Over the years, significant technological advancements have been introduced to improve the quality and quantity of harvests by effectively managing weeds, pests, and diseases. Many studies have focused on identifying plant diseases, as this information aids in making informed decisions about applying fungicides and fertilizers. Advanced systems often employ a combination of image processing and deep learning techniques to identify diseases based on visible symptoms. However, these systems typically rely on pre-existing datasets or images captured in controlled environments. This study showcases the efficacy of utilizing multispectral images captured in visible and Near Infrared (NIR) ranges for identifying plant diseases in real-world environmental conditions. The collected datasets were classified using popular Vision Transformer (ViT) models, including ViT- S16, ViT-BI6, ViT-LI6 and ViT-B32. The results showed impressive training and test accuracies for all the data collected using diverse Kolari vision lenses with 93.71 % and 90.02 %, respectively. This work highlights the potential of utilizing advanced imaging techniques for accurate and reliable plant disease identification in practical field conditions.
- Full Text:
- Date Issued: 2023
- Authors: De Silva, Malitha , Brown, Dane L
- Date: 2023
- Subjects: To be catalogued
- Language: English
- Type: text , article
- Identifier: http://hdl.handle.net/10962/463456 , vital:76410 , xlink:href="https://ieeexplore.ieee.org/abstract/document/10220517"
- Description: Enhancing agricultural practices has become essential in mitigating global hunger. Over the years, significant technological advancements have been introduced to improve the quality and quantity of harvests by effectively managing weeds, pests, and diseases. Many studies have focused on identifying plant diseases, as this information aids in making informed decisions about applying fungicides and fertilizers. Advanced systems often employ a combination of image processing and deep learning techniques to identify diseases based on visible symptoms. However, these systems typically rely on pre-existing datasets or images captured in controlled environments. This study showcases the efficacy of utilizing multispectral images captured in visible and Near Infrared (NIR) ranges for identifying plant diseases in real-world environmental conditions. The collected datasets were classified using popular Vision Transformer (ViT) models, including ViT- S16, ViT-BI6, ViT-LI6 and ViT-B32. The results showed impressive training and test accuracies for all the data collected using diverse Kolari vision lenses with 93.71 % and 90.02 %, respectively. This work highlights the potential of utilizing advanced imaging techniques for accurate and reliable plant disease identification in practical field conditions.
- Full Text:
- Date Issued: 2023
Real-Time Detecting and Tracking of Squids Using YOLOv5
- Kuhlane, Luxolo L, Brown, Dane L, Marais, Marc
- Authors: Kuhlane, Luxolo L , Brown, Dane L , Marais, Marc
- Date: 2023
- Subjects: To be catalogued
- Language: English
- Type: text , article
- Identifier: http://hdl.handle.net/10962/463467 , vital:76411 , xlink:href="https://ieeexplore.ieee.org/abstract/document/10220521"
- Description: This paper proposes a real-time system for detecting and tracking squids using the YOLOv5 object detection algorithm. The system utilizes a large dataset of annotated squid images and videos to train a YOLOv5 model optimized for detecting and tracking squids. The model is fine-tuned to minimize false positives and optimize detection accuracy. The system is deployed on a GPU-enabled device for real-time processing of video streams and tracking of detected squids across frames. The accuracy and speed of the system make it a valuable tool for marine scientists, conservationists, and fishermen to better understand the behavior and distribution of these elusive creatures. Future work includes incorporating additional computer vision techniques and sensor data to improve tracking accuracy and robustness.
- Full Text:
- Date Issued: 2023
- Authors: Kuhlane, Luxolo L , Brown, Dane L , Marais, Marc
- Date: 2023
- Subjects: To be catalogued
- Language: English
- Type: text , article
- Identifier: http://hdl.handle.net/10962/463467 , vital:76411 , xlink:href="https://ieeexplore.ieee.org/abstract/document/10220521"
- Description: This paper proposes a real-time system for detecting and tracking squids using the YOLOv5 object detection algorithm. The system utilizes a large dataset of annotated squid images and videos to train a YOLOv5 model optimized for detecting and tracking squids. The model is fine-tuned to minimize false positives and optimize detection accuracy. The system is deployed on a GPU-enabled device for real-time processing of video streams and tracking of detected squids across frames. The accuracy and speed of the system make it a valuable tool for marine scientists, conservationists, and fishermen to better understand the behavior and distribution of these elusive creatures. Future work includes incorporating additional computer vision techniques and sensor data to improve tracking accuracy and robustness.
- Full Text:
- Date Issued: 2023
Spatiotemporal Convolutions and Video Vision Transformers for Signer-Independent Sign Language Recognition
- Marais, Marc, Brown, Dane L, Connan, James, Boby, Alden
- Authors: Marais, Marc , Brown, Dane L , Connan, James , Boby, Alden
- Date: 2023
- Subjects: To be catalogued
- Language: English
- Type: text , article
- Identifier: http://hdl.handle.net/10962/463478 , vital:76412 , xlink:href="https://ieeexplore.ieee.org/abstract/document/10220534"
- Description: Sign language is a vital tool of communication for individuals who are deaf or hard of hearing. Sign language recognition (SLR) technology can assist in bridging the communication gap between deaf and hearing individuals. However, existing SLR systems are typically signer-dependent, requiring training data from the specific signer for accurate recognition. This presents a significant challenge for practical use, as collecting data from every possible signer is not feasible. This research focuses on developing a signer-independent isolated SLR system to address this challenge. The system implements two model variants on the signer-independent datasets: an R(2+ I)D spatiotemporal convolutional block and a Video Vision transformer. These models learn to extract features from raw sign language videos from the LSA64 dataset and classify signs without needing handcrafted features, explicit segmentation or pose estimation. Overall, the R(2+1)D model architecture significantly outperformed the ViViT architecture for signer-independent SLR on the LSA64 dataset. The R(2+1)D model achieved a near-perfect accuracy of 99.53% on the unseen test set, with the ViViT model yielding an accuracy of 72.19 %. Proving that spatiotemporal convolutions are effective at signer-independent SLR.
- Full Text:
- Date Issued: 2023
- Authors: Marais, Marc , Brown, Dane L , Connan, James , Boby, Alden
- Date: 2023
- Subjects: To be catalogued
- Language: English
- Type: text , article
- Identifier: http://hdl.handle.net/10962/463478 , vital:76412 , xlink:href="https://ieeexplore.ieee.org/abstract/document/10220534"
- Description: Sign language is a vital tool of communication for individuals who are deaf or hard of hearing. Sign language recognition (SLR) technology can assist in bridging the communication gap between deaf and hearing individuals. However, existing SLR systems are typically signer-dependent, requiring training data from the specific signer for accurate recognition. This presents a significant challenge for practical use, as collecting data from every possible signer is not feasible. This research focuses on developing a signer-independent isolated SLR system to address this challenge. The system implements two model variants on the signer-independent datasets: an R(2+ I)D spatiotemporal convolutional block and a Video Vision transformer. These models learn to extract features from raw sign language videos from the LSA64 dataset and classify signs without needing handcrafted features, explicit segmentation or pose estimation. Overall, the R(2+1)D model architecture significantly outperformed the ViViT architecture for signer-independent SLR on the LSA64 dataset. The R(2+1)D model achieved a near-perfect accuracy of 99.53% on the unseen test set, with the ViViT model yielding an accuracy of 72.19 %. Proving that spatiotemporal convolutions are effective at signer-independent SLR.
- Full Text:
- Date Issued: 2023
An evaluation of hand-based algorithms for sign language recognition
- Marais, Marc, Brown, Dane L, Connan, James, Boby, Alden
- Authors: Marais, Marc , Brown, Dane L , Connan, James , Boby, Alden
- Date: 2022
- Subjects: To be catalogued
- Language: English
- Type: text , article
- Identifier: http://hdl.handle.net/10962/465124 , vital:76575 , xlink:href="https://ieeexplore.ieee.org/abstract/document/9856310"
- Description: Sign language recognition is an evolving research field in computer vision, assisting communication between hearing disabled people. Hand gestures contain the majority of the information when signing. Focusing on feature extraction methods to obtain the information stored in hand data in sign language recognition may improve classification accuracy. Pose estimation is a popular method for extracting body and hand landmarks. We implement and compare different feature extraction and segmentation algorithms, focusing on the hands only on the LSA64 dataset. To extract hand landmark coordinates, MediaPipe Holistic is implemented on the sign images. Classification is performed using poplar CNN architectures, namely ResNet and a Pruned VGG network. A separate 1D-CNN is utilised to classify hand landmark coordinates extracted using MediaPipe. The best performance was achieved on the unprocessed raw images using a Pruned VGG network with an accuracy of 95.50%. However, the more computationally efficient model using the hand landmark data and 1D-CNN for classification achieved an accuracy of 94.91%.
- Full Text:
- Date Issued: 2022
- Authors: Marais, Marc , Brown, Dane L , Connan, James , Boby, Alden
- Date: 2022
- Subjects: To be catalogued
- Language: English
- Type: text , article
- Identifier: http://hdl.handle.net/10962/465124 , vital:76575 , xlink:href="https://ieeexplore.ieee.org/abstract/document/9856310"
- Description: Sign language recognition is an evolving research field in computer vision, assisting communication between hearing disabled people. Hand gestures contain the majority of the information when signing. Focusing on feature extraction methods to obtain the information stored in hand data in sign language recognition may improve classification accuracy. Pose estimation is a popular method for extracting body and hand landmarks. We implement and compare different feature extraction and segmentation algorithms, focusing on the hands only on the LSA64 dataset. To extract hand landmark coordinates, MediaPipe Holistic is implemented on the sign images. Classification is performed using poplar CNN architectures, namely ResNet and a Pruned VGG network. A separate 1D-CNN is utilised to classify hand landmark coordinates extracted using MediaPipe. The best performance was achieved on the unprocessed raw images using a Pruned VGG network with an accuracy of 95.50%. However, the more computationally efficient model using the hand landmark data and 1D-CNN for classification achieved an accuracy of 94.91%.
- Full Text:
- Date Issued: 2022
Deep face-iris recognition using robust image segmentation and hyperparameter tuning
- Authors: Brown, Dane L
- Date: 2022
- Subjects: To be catalogued
- Language: English
- Type: text , article
- Identifier: http://hdl.handle.net/10962/465145 , vital:76577 , xlink:href="https://link.springer.com/chapter/10.1007/978-981-16-3728-5_19"
- Description: Biometrics are increasingly being used for tasks that involve sensitive or financial data. Hitherto, security on devices such as smartphones has not been a priority. Furthermore, users tend to ignore the security features in favour of more rapid access to the device. A bimodal system is proposed that enhances security by utilizing face and iris biometrics from a single image. The motivation behind this is the ability to acquire both biometrics simultaneously in one shot. The system’s biometric components: face, iris(es) and their fusion are evaluated. They are also compared to related studies. The best results were yielded by a proposed lightweight Convolutional Neural Network architecture, outperforming tuned VGG-16, Xception, SVM and the related works. The system shows advancements to ‘at-a-distance’ biometric recognition for limited and high computational capacity computing devices. All deep learning algorithms are provided with augmented data, included in the tuning process, enabling additional accuracy gains. Highlights include near-perfect fivefold cross-validation accuracy on the IITD-Iris dataset when performing identification. Verification tests were carried out on the challenging CASIA-Iris-Distance dataset and performed well on few training samples. The proposed system is practical for small or large amounts of training data and shows great promise for at-a-distance recognition and biometric fusion.
- Full Text:
- Date Issued: 2022
- Authors: Brown, Dane L
- Date: 2022
- Subjects: To be catalogued
- Language: English
- Type: text , article
- Identifier: http://hdl.handle.net/10962/465145 , vital:76577 , xlink:href="https://link.springer.com/chapter/10.1007/978-981-16-3728-5_19"
- Description: Biometrics are increasingly being used for tasks that involve sensitive or financial data. Hitherto, security on devices such as smartphones has not been a priority. Furthermore, users tend to ignore the security features in favour of more rapid access to the device. A bimodal system is proposed that enhances security by utilizing face and iris biometrics from a single image. The motivation behind this is the ability to acquire both biometrics simultaneously in one shot. The system’s biometric components: face, iris(es) and their fusion are evaluated. They are also compared to related studies. The best results were yielded by a proposed lightweight Convolutional Neural Network architecture, outperforming tuned VGG-16, Xception, SVM and the related works. The system shows advancements to ‘at-a-distance’ biometric recognition for limited and high computational capacity computing devices. All deep learning algorithms are provided with augmented data, included in the tuning process, enabling additional accuracy gains. Highlights include near-perfect fivefold cross-validation accuracy on the IITD-Iris dataset when performing identification. Verification tests were carried out on the challenging CASIA-Iris-Distance dataset and performed well on few training samples. The proposed system is practical for small or large amounts of training data and shows great promise for at-a-distance recognition and biometric fusion.
- Full Text:
- Date Issued: 2022
Deep Learning Approach to Image Deblurring and Image Super-Resolution using DeblurGAN and SRGAN
- Kuhlane, Luxolo L, Brown, Dane L, Connan, James, Boby, Alden, Marais, Marc
- 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
- 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
Deep Palmprint Recognition with Alignment and Augmentation of Limited Training Samples
- Brown, Dane L, Bradshaw, Karen L
- Authors: Brown, Dane L , Bradshaw, Karen L
- Date: 2022
- Subjects: To be catalogued
- Language: English
- Type: text , article
- Identifier: http://hdl.handle.net/10962/440249 , vital:73760 , xlink:href="https://doi.org/10.1007/s42979-021-00859-3"
- Description: This paper builds upon a previously proposed automatic palmprint alignment and classification system. The proposed system was geared towards palmprints acquired from either contact or contactless sensors. It was robust to finger location and fist shape changes—accurately extracting the palmprints in images without fingers. An extension to this previous work includes comparisons of traditional and deep learning models, both with hyperparameter tuning. The proposed methods are compared with related verification systems and a detailed evaluation of open-set identification. The best results were yielded by a proposed Convolutional Neural Network, based on VGG-16, and outperforming tuned VGG-16 and Xception architectures. All deep learning algorithms are provided with augmented data, included in the tuning process, enabling significant accuracy gains. Highlights include near-zero and zero EER on IITD-Palmprint verification using one training sample and leave-one-out strategy, respectively. Therefore, the proposed palmprint system is practical as it is effective on data containing many and few training examples.
- Full Text:
- Date Issued: 2022
- Authors: Brown, Dane L , Bradshaw, Karen L
- Date: 2022
- Subjects: To be catalogued
- Language: English
- Type: text , article
- Identifier: http://hdl.handle.net/10962/440249 , vital:73760 , xlink:href="https://doi.org/10.1007/s42979-021-00859-3"
- Description: This paper builds upon a previously proposed automatic palmprint alignment and classification system. The proposed system was geared towards palmprints acquired from either contact or contactless sensors. It was robust to finger location and fist shape changes—accurately extracting the palmprints in images without fingers. An extension to this previous work includes comparisons of traditional and deep learning models, both with hyperparameter tuning. The proposed methods are compared with related verification systems and a detailed evaluation of open-set identification. The best results were yielded by a proposed Convolutional Neural Network, based on VGG-16, and outperforming tuned VGG-16 and Xception architectures. All deep learning algorithms are provided with augmented data, included in the tuning process, enabling significant accuracy gains. Highlights include near-zero and zero EER on IITD-Palmprint verification using one training sample and leave-one-out strategy, respectively. Therefore, the proposed palmprint system is practical as it is effective on data containing many and few training examples.
- Full Text:
- Date Issued: 2022
Deep palmprint recognition with alignment and augmentation of limited training samples
- Brown, Dane L, Bradshaw, Karen L
- Authors: Brown, Dane L , Bradshaw, Karen L
- Date: 2022
- Subjects: To be catalogued
- Language: English
- Type: text , article
- Identifier: http://hdl.handle.net/10962/464074 , vital:76473 , xlink:href="https://doi.org/10.1007/s42979-021-00859-3"
- Description: This paper builds upon a previously proposed automatic palmprint alignment and classification system. The proposed system was geared towards palmprints acquired from either contact or contactless sensors. It was robust to finger location and fist shape changes—accurately extracting the palmprints in images without fingers. An extension to this previous work includes comparisons of traditional and deep learning models, both with hyperparameter tuning. The proposed methods are compared with related verification systems and a detailed evaluation of open-set identification. The best results were yielded by a proposed Convolutional Neural Network, based on VGG-16, and outperforming tuned VGG-16 and Xception architectures. All deep learning algorithms are provided with augmented data, included in the tuning process, enabling significant accuracy gains. Highlights include near-zero and zero EER on IITD-Palmprint verification using one training sample and leave-one-out strategy, respectively. Therefore, the proposed palmprint system is practical as it is effective on data containing many and few training examples.
- Full Text:
- Date Issued: 2022
- Authors: Brown, Dane L , Bradshaw, Karen L
- Date: 2022
- Subjects: To be catalogued
- Language: English
- Type: text , article
- Identifier: http://hdl.handle.net/10962/464074 , vital:76473 , xlink:href="https://doi.org/10.1007/s42979-021-00859-3"
- Description: This paper builds upon a previously proposed automatic palmprint alignment and classification system. The proposed system was geared towards palmprints acquired from either contact or contactless sensors. It was robust to finger location and fist shape changes—accurately extracting the palmprints in images without fingers. An extension to this previous work includes comparisons of traditional and deep learning models, both with hyperparameter tuning. The proposed methods are compared with related verification systems and a detailed evaluation of open-set identification. The best results were yielded by a proposed Convolutional Neural Network, based on VGG-16, and outperforming tuned VGG-16 and Xception architectures. All deep learning algorithms are provided with augmented data, included in the tuning process, enabling significant accuracy gains. Highlights include near-zero and zero EER on IITD-Palmprint verification using one training sample and leave-one-out strategy, respectively. Therefore, the proposed palmprint system is practical as it is effective on data containing many and few training examples.
- Full Text:
- Date Issued: 2022
Exploring the Incremental Improvements of YOLOv7 over YOLOv5 for Character Recognition
- Boby, Alden, Brown, Dane L, Connan, James, Marais, Marc
- Authors: Boby, Alden , Brown, Dane L , Connan, James , Marais, Marc
- Date: 2022
- Subjects: To be catalogued
- Language: English
- Type: text , article
- Identifier: http://hdl.handle.net/10962/463395 , vital:76405 , xlink:href="https://link.springer.com/chapter/10.1007/978-3-031-35644-5_5"
- Description: Technological advances are being applied to aspects of life to improve quality of living and efficiency. This speaks specifically to automation, especially in the industry. The growing number of vehicles on the road has presented a need to monitor more vehicles than ever to enforce traffic rules. One way to identify a vehicle is through its licence plate, which contains a unique string of characters that make it identifiable within an external database. Detecting characters on a licence plate using an object detector has only recently been explored. This paper uses the latest versions of the YOLO object detector to perform character recognition on licence plate images. This paper expands upon existing object detection-based character recognition by investigating how improvements in the framework translate to licence plate character recognition accuracy compared to character recognition based on older architectures. Results from this paper indicate that the newer YOLO models have increased performance over older YOLO-based character recognition models such as CRNET.
- Full Text:
- Date Issued: 2022
- Authors: Boby, Alden , Brown, Dane L , Connan, James , Marais, Marc
- Date: 2022
- Subjects: To be catalogued
- Language: English
- Type: text , article
- Identifier: http://hdl.handle.net/10962/463395 , vital:76405 , xlink:href="https://link.springer.com/chapter/10.1007/978-3-031-35644-5_5"
- Description: Technological advances are being applied to aspects of life to improve quality of living and efficiency. This speaks specifically to automation, especially in the industry. The growing number of vehicles on the road has presented a need to monitor more vehicles than ever to enforce traffic rules. One way to identify a vehicle is through its licence plate, which contains a unique string of characters that make it identifiable within an external database. Detecting characters on a licence plate using an object detector has only recently been explored. This paper uses the latest versions of the YOLO object detector to perform character recognition on licence plate images. This paper expands upon existing object detection-based character recognition by investigating how improvements in the framework translate to licence plate character recognition accuracy compared to character recognition based on older architectures. Results from this paper indicate that the newer YOLO models have increased performance over older YOLO-based character recognition models such as CRNET.
- Full Text:
- Date Issued: 2022