In this work, deep CNN based model have been suggested for face recognition. CNN is employed to extract unique facial features and softmax classifier is applied to classify facial images in a fully connected layer of CNN. The experiments conducted in Extended YALE B and FERET databases for smaller batch sizes and low value of learning rate, showed that the proposed model has improved the face recognition accuracy. Accuracy rates of up to 96.2% is achieved using the proposed model in Extended Yale B database. To improve the accuracy rate further, preprocessing techniques like SQI, HE, LTISN, GIC and DoG are applied to the CNN model. After the application of preprocessing techniques, the improved accuracy of 99.8% is achieved with deep CNN model for the YALE B Extended Database. In FERET Database with frontal face, before the application of preprocessing techniques, CNN model yields the maximum accuracy of 71.4%. After applying the above-mentioned preprocessing techniques, the accuracy is improved to 76.3%.
A NOVEL SCHEME FOR ACCURATE REMAINING USEFUL LIFE PREDICTION FOR INDUSTRIAL I...ijaia
In the era of the fourth industrial revolution, measuring and ensuring the reliability, efficiency and safety of the industrial systems and components are one of the uppermost key concern. In addition, predicting performance degradation or remaining useful life (RUL) of an equipment over time based on its historical sensor data enables companies to greatly reduce their maintenance cost. In this way, companies can prevent costly unexpected breakdown and become more profitable and competitive in the marketplace. This paper introduces a deep learning-based method by combining CNN(Convolutional Neural Networks) and LSTM (Long Short-Term Memory)neural networks to predict RUL for industrial equipment. The proposed method does not depend upon any degradation trend assumptions and it can learn complex temporal representative and distinguishing patterns in the sensor data. In order to evaluate the efficiency and effectiveness of the proposed method, we evaluated it on two different experiment: RUL estimation and predicting the status of the IoT devices in 2-week period. Experiments are conducted on a publicly available NASA’s turbo fan-engine dataset. Based on the experiment results, the deep learning-based approach achieved high prediction accuracy. Moreover, the results show that the method outperforms standard well-accepted machine learning algorithms and accomplishes competitive performance when compared to the state-of-the art methods
FACIAL AGE ESTIMATION USING TRANSFER LEARNING AND BAYESIAN OPTIMIZATION BASED...sipij
Age estimation of unrestricted imaging circumstances has attracted an augmented recognition as it is
appropriate in several real-world applications such as surveillance, face recognition, age synthesis, access
control, and electronic customer relationship management. Current deep learning-based methods have
displayed encouraging performance in age estimation field. Males and Females have a variable type of
appearance aging pattern; this results in age differently. This fact leads to assuming that using gender
information may improve the age estimator performance. We have proposed a novel model based on
Gender Classification. A Convolutional Neural Network (CNN) is used to get Gender Information, then
Bayesian Optimization is applied to this pre-trained CNN when fine-tuned for age estimation task.
Bayesian Optimization reduces the classification error on the validation set for the pre-trained model.
Extensive experiments are done to assess our proposed model on two data sets: FERET and FG-NET. The
experiments’ result indicates that using a pre-trained CNN containing Gender Information with Bayesian
Optimization outperforms the state of the arts on FERET and FG-NET data sets with a Mean Absolute
Error (MAE) of 1.2 and 2.67 respectively.
IJRET : International Journal of Research in Engineering and Technology is an international peer reviewed, online journal published by eSAT Publishing House for the enhancement of research in various disciplines of Engineering and Technology. The aim and scope of the journal is to provide an academic medium and an important reference for the advancement and dissemination of research results that support high-level learning, teaching and research in the fields of Engineering and Technology. We bring together Scientists, Academician, Field Engineers, Scholars and Students of related fields of Engineering and Technology
Content Based Image Retrieval for Unlabelled ImagesIOSR Journals
Abstract: Recently, content-based image retrieval has become hot topic and the techniques of content-based
image retrieval have been achieved good development. Content-based image retrieval systems were introduced
to address the problems associated with text-based image retrieval. In this paper, basic components of contentbased
image retrieval system are introduced here. Images are classified as lablled and unlablled images. Here
survey on content based image retrieval given with some Image retrieval methods based on unlabelled data like
D-EM, SVM, Relevance Feedback, Semi-Supervised/Active Learning, Transductive Learning, Bootstrapping
SVM, Active learning, SSMIL and Label propagation Methods are presented in this paper. Comparison of these
all methods is also presented in this paper.
Keywords: Image retrieval, CBIR, Unlabelled images, SVM
A NOVEL SCHEME FOR ACCURATE REMAINING USEFUL LIFE PREDICTION FOR INDUSTRIAL I...ijaia
In the era of the fourth industrial revolution, measuring and ensuring the reliability, efficiency and safety of the industrial systems and components are one of the uppermost key concern. In addition, predicting performance degradation or remaining useful life (RUL) of an equipment over time based on its historical sensor data enables companies to greatly reduce their maintenance cost. In this way, companies can prevent costly unexpected breakdown and become more profitable and competitive in the marketplace. This paper introduces a deep learning-based method by combining CNN(Convolutional Neural Networks) and LSTM (Long Short-Term Memory)neural networks to predict RUL for industrial equipment. The proposed method does not depend upon any degradation trend assumptions and it can learn complex temporal representative and distinguishing patterns in the sensor data. In order to evaluate the efficiency and effectiveness of the proposed method, we evaluated it on two different experiment: RUL estimation and predicting the status of the IoT devices in 2-week period. Experiments are conducted on a publicly available NASA’s turbo fan-engine dataset. Based on the experiment results, the deep learning-based approach achieved high prediction accuracy. Moreover, the results show that the method outperforms standard well-accepted machine learning algorithms and accomplishes competitive performance when compared to the state-of-the art methods
FACIAL AGE ESTIMATION USING TRANSFER LEARNING AND BAYESIAN OPTIMIZATION BASED...sipij
Age estimation of unrestricted imaging circumstances has attracted an augmented recognition as it is
appropriate in several real-world applications such as surveillance, face recognition, age synthesis, access
control, and electronic customer relationship management. Current deep learning-based methods have
displayed encouraging performance in age estimation field. Males and Females have a variable type of
appearance aging pattern; this results in age differently. This fact leads to assuming that using gender
information may improve the age estimator performance. We have proposed a novel model based on
Gender Classification. A Convolutional Neural Network (CNN) is used to get Gender Information, then
Bayesian Optimization is applied to this pre-trained CNN when fine-tuned for age estimation task.
Bayesian Optimization reduces the classification error on the validation set for the pre-trained model.
Extensive experiments are done to assess our proposed model on two data sets: FERET and FG-NET. The
experiments’ result indicates that using a pre-trained CNN containing Gender Information with Bayesian
Optimization outperforms the state of the arts on FERET and FG-NET data sets with a Mean Absolute
Error (MAE) of 1.2 and 2.67 respectively.
IJRET : International Journal of Research in Engineering and Technology is an international peer reviewed, online journal published by eSAT Publishing House for the enhancement of research in various disciplines of Engineering and Technology. The aim and scope of the journal is to provide an academic medium and an important reference for the advancement and dissemination of research results that support high-level learning, teaching and research in the fields of Engineering and Technology. We bring together Scientists, Academician, Field Engineers, Scholars and Students of related fields of Engineering and Technology
Content Based Image Retrieval for Unlabelled ImagesIOSR Journals
Abstract: Recently, content-based image retrieval has become hot topic and the techniques of content-based
image retrieval have been achieved good development. Content-based image retrieval systems were introduced
to address the problems associated with text-based image retrieval. In this paper, basic components of contentbased
image retrieval system are introduced here. Images are classified as lablled and unlablled images. Here
survey on content based image retrieval given with some Image retrieval methods based on unlabelled data like
D-EM, SVM, Relevance Feedback, Semi-Supervised/Active Learning, Transductive Learning, Bootstrapping
SVM, Active learning, SSMIL and Label propagation Methods are presented in this paper. Comparison of these
all methods is also presented in this paper.
Keywords: Image retrieval, CBIR, Unlabelled images, SVM
DATA AUGMENTATION TECHNIQUES AND TRANSFER LEARNING APPROACHES APPLIED TO FACI...ijaia
The face expression is the first thing we pay attention to when we want to understand a person’s state of
mind. Thus, the ability to recognize facial expressions in an automatic way is a very interesting research
field. In this paper, because the small size of available training datasets, we propose a novel data
augmentation technique that improves the performances in the recognition task. We apply geometrical
transformations and build from scratch GAN models able to generate new synthetic images for each
emotion type. Thus, on the augmented datasets we fine tune pretrained convolutional neural networks with
different architectures. To measure the generalization ability of the models, we apply extra-database
protocol approach, namely we train models on the augmented versions of training dataset and test them on
two different databases. The combination of these techniques allows to reach average accuracy values of
the order of 85% for the InceptionResNetV2 model.
Segmentation of Brain MR Images for Tumor Extraction by Combining Kmeans Clus...CSCJournals
Segmentation of images holds an important position in the area of image processing. It becomes more important while typically dealing with medical images where pre-surgery and post surgery decisions are required for the purpose of initiating and speeding up the recovery process [5] Computer aided detection of abnormal growth of tissues is primarily motivated by the necessity of achieving maximum possible accuracy. Manual segmentation of these abnormal tissues cannot be compared with modern day’s high speed computing machines which enable us to visually observe the volume and location of unwanted tissues. A well known segmentation problem within MRI is the task of labeling voxels according to their tissue type which include White Matter (WM), Grey Matter (GM) , Cerebrospinal Fluid (CSF) and sometimes pathological tissues like tumor etc. This paper describes an efficient method for automatic brain tumor segmentation for the extraction of tumor tissues from MR images. It combines Perona and Malik anisotropic diffusion model for image enhancement and Kmeans clustering technique for grouping tissues belonging to a specific group. The proposed method uses T1, T2 and PD weighted gray level intensity images. The proposed technique produced appreciative results
Dataset Pre-Processing and Artificial Augmentation, Network Architecture and ...INFOGAIN PUBLICATION
Training a Convolutional Neural Network (CNN) based classifier is dependent on a large number of factors. These factors involve tasks such as aggregation of apt dataset, arriving at a suitable CNN network, processing of the dataset, and selecting the training parameters to arrive at the desired classification results. This review includes pre-processing techniques and dataset augmentation techniques used in various CNN based classification researches. In many classification problems, it is usually observed that the quality of dataset is responsible for proper training of CNN network, and this quality is judged on the basis of variations in data for every class. It is not usual to find such a pre-made dataset due to many natural concerns. Also it is recommended to have a large dataset, which is again not usually made available directly as a dataset. In some cases, the noise present in the dataset may not prove useful for training, while in others, researchers prefer to add noise to certain images to make the network less vulnerable to unwanted variations. Hence, researchers use artificial digital imaging techniques to derive variations in the dataset and clear or add noise. Thus, the presented paper accumulates state-of-the-art works that used the pre-processing and artificial augmentation of dataset before training. The next part to data augmentation is training, which includes proper selection of several parameters and a suitable CNN architecture. This paper also includes such network characteristics, dataset characteristics and training methodologies used in biomedical imaging, vision modules of autonomous driverless cars, and a few general vision based applications.
Identification and classification of citrus fruit
more precisely and economically under natural
illumination circumstances. The goal of this paper was
to develop a robust and swift algorithm to detect and
categorize citrus fruit with different fruit dimensions
and under different illumination conditions. To identify
object residing in image, the image has to be described
or represented by certain features. In this paper,
proposed a Image pre-processing and stem removal
process for deriving the citrus fruit classification. The
image pre-processing is carried out using RGB to HSV
color model conversion and noise removal using
Gaussian method and Stem removal extraction of using
morphological open image, distance, top-hat filtering
and Finest StemcompleteRm gray level thresholding.
For stem removal, difference stages representation and
description techniques are discuss in this paper. Stem
removal techniques play an important role in systems
for correct object recognition, matching, extracting, and
analysis.
MITIGATION TECHNIQUES TO OVERCOME DATA HARM IN MODEL BUILDING FOR MLijaia
Given the impact of Machine Learning (ML) on individuals and the society, understanding how harm might
be occur throughout the ML life cycle becomes critical more than ever. By offering a framework to
determine distinct potential sources of downstream harm in ML pipeline, the paper demonstrates the
importance of choices throughout distinct phases of data collection, development, and deployment that
extend far beyond just model training. Relevant mitigation techniques are also suggested for being used
instead of merely relying on generic notions of what counts as fairness.
DEEP-LEARNING-BASED HUMAN INTENTION PREDICTION WITH DATA AUGMENTATIONijaia
Data augmentation has been broadly applied in training deep-learning models to increase the diversity of
data. This study ingestigates the effectiveness of different data augmentation methods for deep-learningbased human intention prediction when only limited training data is available. A human participant pitches
a ball to nine potential targets in our experiment. We expect to predict which target the participant pitches
the ball to. Firstly, the effectiveness of 10 data augmentation groups is evaluated on a single-participant
data set using RGB images. Secondly, the best data augmentation method (i.e., random cropping) on the
single-participant data set is further evaluated on a multi-participant data set to assess its generalization
ability. Finally, the effectiveness of random cropping on fusion data of RGB images and optical flow is
evaluated on both single- and multi-participant data sets. Experiment results show that: 1) Data
augmentation methods that crop or deform images can improve the prediction performance; 2) Random
cropping can be generalized to the multi-participant data set (prediction accuracy is improved from 50%
to 57.4%); and 3) Random cropping with fusion data of RGB images and optical flow can further improve
the prediction accuracy from 57.4% to 63.9% on the multi-participant data set.
REVIEWING PROCESS MINING APPLICATIONS AND TECHNIQUES IN EDUCATIONijaia
Process Mining (PM) emerged from business process management but has recently been applied to
educational data and has been found to facilitate the understanding of the educational process.
Educational Process Mining (EPM) bridges the gap between process analysis and data analysis, based on
the techniques of model discovery, conformance checking and extension of existing process models. We
present a systematic review of the recent and current status of research in the EPM domain, focusing on
application domains, techniques, tools and models, to highlight the use of EPM in comprehending and
improving educational processes.
IMBALANCED DATASET EFFECT ON CNN-BASED CLASSIFIER PERFORMANCE FOR FACE RECOGN...gerogepatton
Facial Recognition is integral to numerous modern applications, such as security systems, social media
platforms, and augmented reality apps. The success of these systems heavily depends on the performance
of the Face Recognition models they use, specifically Convolutional Neural Networks (CNNs). However,
many real-world classification tasks encounter imbalanced datasets, with some classes significantly
underrepresented. Face Recognition models that do not address this class imbalance tend to exhibit poor
performance, especially in tasks involving a wide range of faces to identify (multi-class problems). This
research examines how class imbalance in datasets impacts the creation of neural network classifiers for
Facial Recognition. Initially, we crafted a Convolutional Neural Network model for facial recognition,
integrating hybrid resampling methods (oversampling and under-sampling) to address dataset imbalances.
In addition, augmentation techniques were implemented to enhance generalization capabilities and overall
performance. Through comprehensive experimentation, we assess the influence of imbalanced datasets on
the performance of the CNN-based classifier. Using Pins face data, we conducted an empirical study,
evaluating conclusions based on accuracy, precision, recall, and F1-score measurements. A comparative
analysis demonstrates that the performance of the proposed Convolutional Neural Network classifier
diminishes in the presence of dataset class imbalances. Conversely, the proposed system, utilizing data
resampling techniques, notably enhances classification performance for imbalanced datasets. This study
underscores the efficacy of data resampling approaches in augmenting the performance of Face
Recognition models, presenting prospects for more dependable and efficient future systems.
Imbalanced Dataset Effect on CNN-Based Classifier Performance for Face Recogn...gerogepatton
Facial Recognition is integral to numerous modern applications, such as security systems, social media
platforms, and augmented reality apps. The success of these systems heavily depends on the performance
of the Face Recognition models they use, specifically Convolutional Neural Networks (CNNs). However,
many real-world classification tasks encounter imbalanced datasets, with some classes significantly
underrepresented. Face Recognition models that do not address this class imbalance tend to exhibit poor
performance, especially in tasks involving a wide range of faces to identify (multi-class problems). This
research examines how class imbalance in datasets impacts the creation of neural network classifiers for
Facial Recognition. Initially, we crafted a Convolutional Neural Network model for facial recognition,
integrating hybrid resampling methods (oversampling and under-sampling) to address dataset imbalances.
In addition, augmentation techniques were implemented to enhance generalization capabilities and overall
performance. Through comprehensive experimentation, we assess the influence of imbalanced datasets on
the performance of the CNN-based classifier. Using Pins face data, we conducted an empirical study,
evaluating conclusions based on accuracy, precision, recall, and F1-score measurements. A comparative
analysis demonstrates that the performance of the proposed Convolutional Neural Network classifier
diminishes in the presence of dataset class imbalances. Conversely, the proposed system, utilizing data
resampling techniques, notably enhances classification performance for imbalanced datasets. This study
underscores the efficacy of data resampling approaches in augmenting the performance of Face
Recognition models, presenting prospects for more dependable and efficient future systems.
DATA AUGMENTATION TECHNIQUES AND TRANSFER LEARNING APPROACHES APPLIED TO FACI...ijaia
The face expression is the first thing we pay attention to when we want to understand a person’s state of
mind. Thus, the ability to recognize facial expressions in an automatic way is a very interesting research
field. In this paper, because the small size of available training datasets, we propose a novel data
augmentation technique that improves the performances in the recognition task. We apply geometrical
transformations and build from scratch GAN models able to generate new synthetic images for each
emotion type. Thus, on the augmented datasets we fine tune pretrained convolutional neural networks with
different architectures. To measure the generalization ability of the models, we apply extra-database
protocol approach, namely we train models on the augmented versions of training dataset and test them on
two different databases. The combination of these techniques allows to reach average accuracy values of
the order of 85% for the InceptionResNetV2 model.
Segmentation of Brain MR Images for Tumor Extraction by Combining Kmeans Clus...CSCJournals
Segmentation of images holds an important position in the area of image processing. It becomes more important while typically dealing with medical images where pre-surgery and post surgery decisions are required for the purpose of initiating and speeding up the recovery process [5] Computer aided detection of abnormal growth of tissues is primarily motivated by the necessity of achieving maximum possible accuracy. Manual segmentation of these abnormal tissues cannot be compared with modern day’s high speed computing machines which enable us to visually observe the volume and location of unwanted tissues. A well known segmentation problem within MRI is the task of labeling voxels according to their tissue type which include White Matter (WM), Grey Matter (GM) , Cerebrospinal Fluid (CSF) and sometimes pathological tissues like tumor etc. This paper describes an efficient method for automatic brain tumor segmentation for the extraction of tumor tissues from MR images. It combines Perona and Malik anisotropic diffusion model for image enhancement and Kmeans clustering technique for grouping tissues belonging to a specific group. The proposed method uses T1, T2 and PD weighted gray level intensity images. The proposed technique produced appreciative results
Dataset Pre-Processing and Artificial Augmentation, Network Architecture and ...INFOGAIN PUBLICATION
Training a Convolutional Neural Network (CNN) based classifier is dependent on a large number of factors. These factors involve tasks such as aggregation of apt dataset, arriving at a suitable CNN network, processing of the dataset, and selecting the training parameters to arrive at the desired classification results. This review includes pre-processing techniques and dataset augmentation techniques used in various CNN based classification researches. In many classification problems, it is usually observed that the quality of dataset is responsible for proper training of CNN network, and this quality is judged on the basis of variations in data for every class. It is not usual to find such a pre-made dataset due to many natural concerns. Also it is recommended to have a large dataset, which is again not usually made available directly as a dataset. In some cases, the noise present in the dataset may not prove useful for training, while in others, researchers prefer to add noise to certain images to make the network less vulnerable to unwanted variations. Hence, researchers use artificial digital imaging techniques to derive variations in the dataset and clear or add noise. Thus, the presented paper accumulates state-of-the-art works that used the pre-processing and artificial augmentation of dataset before training. The next part to data augmentation is training, which includes proper selection of several parameters and a suitable CNN architecture. This paper also includes such network characteristics, dataset characteristics and training methodologies used in biomedical imaging, vision modules of autonomous driverless cars, and a few general vision based applications.
Identification and classification of citrus fruit
more precisely and economically under natural
illumination circumstances. The goal of this paper was
to develop a robust and swift algorithm to detect and
categorize citrus fruit with different fruit dimensions
and under different illumination conditions. To identify
object residing in image, the image has to be described
or represented by certain features. In this paper,
proposed a Image pre-processing and stem removal
process for deriving the citrus fruit classification. The
image pre-processing is carried out using RGB to HSV
color model conversion and noise removal using
Gaussian method and Stem removal extraction of using
morphological open image, distance, top-hat filtering
and Finest StemcompleteRm gray level thresholding.
For stem removal, difference stages representation and
description techniques are discuss in this paper. Stem
removal techniques play an important role in systems
for correct object recognition, matching, extracting, and
analysis.
MITIGATION TECHNIQUES TO OVERCOME DATA HARM IN MODEL BUILDING FOR MLijaia
Given the impact of Machine Learning (ML) on individuals and the society, understanding how harm might
be occur throughout the ML life cycle becomes critical more than ever. By offering a framework to
determine distinct potential sources of downstream harm in ML pipeline, the paper demonstrates the
importance of choices throughout distinct phases of data collection, development, and deployment that
extend far beyond just model training. Relevant mitigation techniques are also suggested for being used
instead of merely relying on generic notions of what counts as fairness.
DEEP-LEARNING-BASED HUMAN INTENTION PREDICTION WITH DATA AUGMENTATIONijaia
Data augmentation has been broadly applied in training deep-learning models to increase the diversity of
data. This study ingestigates the effectiveness of different data augmentation methods for deep-learningbased human intention prediction when only limited training data is available. A human participant pitches
a ball to nine potential targets in our experiment. We expect to predict which target the participant pitches
the ball to. Firstly, the effectiveness of 10 data augmentation groups is evaluated on a single-participant
data set using RGB images. Secondly, the best data augmentation method (i.e., random cropping) on the
single-participant data set is further evaluated on a multi-participant data set to assess its generalization
ability. Finally, the effectiveness of random cropping on fusion data of RGB images and optical flow is
evaluated on both single- and multi-participant data sets. Experiment results show that: 1) Data
augmentation methods that crop or deform images can improve the prediction performance; 2) Random
cropping can be generalized to the multi-participant data set (prediction accuracy is improved from 50%
to 57.4%); and 3) Random cropping with fusion data of RGB images and optical flow can further improve
the prediction accuracy from 57.4% to 63.9% on the multi-participant data set.
REVIEWING PROCESS MINING APPLICATIONS AND TECHNIQUES IN EDUCATIONijaia
Process Mining (PM) emerged from business process management but has recently been applied to
educational data and has been found to facilitate the understanding of the educational process.
Educational Process Mining (EPM) bridges the gap between process analysis and data analysis, based on
the techniques of model discovery, conformance checking and extension of existing process models. We
present a systematic review of the recent and current status of research in the EPM domain, focusing on
application domains, techniques, tools and models, to highlight the use of EPM in comprehending and
improving educational processes.
IMBALANCED DATASET EFFECT ON CNN-BASED CLASSIFIER PERFORMANCE FOR FACE RECOGN...gerogepatton
Facial Recognition is integral to numerous modern applications, such as security systems, social media
platforms, and augmented reality apps. The success of these systems heavily depends on the performance
of the Face Recognition models they use, specifically Convolutional Neural Networks (CNNs). However,
many real-world classification tasks encounter imbalanced datasets, with some classes significantly
underrepresented. Face Recognition models that do not address this class imbalance tend to exhibit poor
performance, especially in tasks involving a wide range of faces to identify (multi-class problems). This
research examines how class imbalance in datasets impacts the creation of neural network classifiers for
Facial Recognition. Initially, we crafted a Convolutional Neural Network model for facial recognition,
integrating hybrid resampling methods (oversampling and under-sampling) to address dataset imbalances.
In addition, augmentation techniques were implemented to enhance generalization capabilities and overall
performance. Through comprehensive experimentation, we assess the influence of imbalanced datasets on
the performance of the CNN-based classifier. Using Pins face data, we conducted an empirical study,
evaluating conclusions based on accuracy, precision, recall, and F1-score measurements. A comparative
analysis demonstrates that the performance of the proposed Convolutional Neural Network classifier
diminishes in the presence of dataset class imbalances. Conversely, the proposed system, utilizing data
resampling techniques, notably enhances classification performance for imbalanced datasets. This study
underscores the efficacy of data resampling approaches in augmenting the performance of Face
Recognition models, presenting prospects for more dependable and efficient future systems.
Imbalanced Dataset Effect on CNN-Based Classifier Performance for Face Recogn...gerogepatton
Facial Recognition is integral to numerous modern applications, such as security systems, social media
platforms, and augmented reality apps. The success of these systems heavily depends on the performance
of the Face Recognition models they use, specifically Convolutional Neural Networks (CNNs). However,
many real-world classification tasks encounter imbalanced datasets, with some classes significantly
underrepresented. Face Recognition models that do not address this class imbalance tend to exhibit poor
performance, especially in tasks involving a wide range of faces to identify (multi-class problems). This
research examines how class imbalance in datasets impacts the creation of neural network classifiers for
Facial Recognition. Initially, we crafted a Convolutional Neural Network model for facial recognition,
integrating hybrid resampling methods (oversampling and under-sampling) to address dataset imbalances.
In addition, augmentation techniques were implemented to enhance generalization capabilities and overall
performance. Through comprehensive experimentation, we assess the influence of imbalanced datasets on
the performance of the CNN-based classifier. Using Pins face data, we conducted an empirical study,
evaluating conclusions based on accuracy, precision, recall, and F1-score measurements. A comparative
analysis demonstrates that the performance of the proposed Convolutional Neural Network classifier
diminishes in the presence of dataset class imbalances. Conversely, the proposed system, utilizing data
resampling techniques, notably enhances classification performance for imbalanced datasets. This study
underscores the efficacy of data resampling approaches in augmenting the performance of Face
Recognition models, presenting prospects for more dependable and efficient future systems.
A NOVEL SCHEME FOR ACCURATE REMAINING USEFUL LIFE PREDICTION FOR INDUSTRIAL I...gerogepatton
In the era of the fourth industrial revolution, measuring and ensuring the reliability, efficiency and safety of the industrial systems and components are one of the uppermost key concern. In addition, predicting performance degradation or remaining useful life (RUL) of an equipment over time based on its historical sensor data enables companies to greatly reduce their maintenance cost. In this way, companies can prevent costly unexpected breakdown and become more profitable and competitive in the marketplace. This paper introduces a deep learning-based method by combining CNN(Convolutional Neural Networks) and LSTM (Long Short-Term Memory)neural networks to predict RUL for industrial equipment. The proposed method does not depend upon any degradation trend assumptions and it can learn complex temporal representative and distinguishing patterns in the sensor data. In order to evaluate the efficiency and effectiveness of the proposed method, we evaluated it on two different experiment: RUL estimation and predicting the status of the IoT devices in 2-week period. Experiments are conducted on a publicly available NASA’s turbo fan-engine dataset. Based on the experiment results, the deep learning-based approach achieved high prediction accuracy. Moreover, the results show that the method outperforms standard well-accepted machine learning algorithms and accomplishes competitive performance when compared to the state-of-the art methods
Paper Explained: RandAugment: Practical automated data augmentation with a re...Devansh16
RandAugment: Practical automated data augmentation with a reduced search space is a paper that proposes a new Data Augmentation technique that outperforms all current techniques while being cheaper.
A hybrid approach for face recognition using a convolutional neural network c...IAESIJAI
Facial recognition technology has been used in many fields such as security,
biometric identification, robotics, video surveillance, health, and commerce
due to its ease of implementation and minimal data processing time.
However, this technology is influenced by the presence of variations such as
pose, lighting, or occlusion. In this paper, we propose a new approach to
improve the accuracy rate of face recognition in the presence of variation or
occlusion, by combining feature extraction with a histogram of oriented
gradient (HOG), scale invariant feature transform (SIFT), Gabor, and the
Canny contour detector techniques, as well as a convolutional neural
network (CNN) architecture, tested with several combinations of the
activation function used (Softmax and Segmoïd) and the optimization
algorithm used during training (adam, Adamax, RMSprop, and stochastic
gradient descent (SGD)). For this, a preprocessing was performed on two
databases of our database of faces (ORL) and Sheffield faces used, then we
perform a feature extraction operation with the mentioned techniques and
then pass them to our used CNN architecture. The results of our simulations
show a high performance of the SIFT+CNN combination, in the case of the
presence of variations with an accuracy rate up to 100%.
End-to-end deep auto-encoder for segmenting a moving object with limited tra...IJECEIAES
Deep learning-based approaches have been widely used in various applications, including segmentation and classification. However, a large amount of data is required to train such techniques. Indeed, in the surveillance video domain, there are few accessible data due to acquisition and experiment complexity. In this paper, we propose an end-to-end deep auto-encoder system for object segmenting from surveillance videos. Our main purpose is to enhance the process of distinguishing the foreground object when only limited data are available. To this end, we propose two approaches based on transfer learning and multi-depth auto-encoders to avoid over-fitting by combining classical data augmentation and principal component analysis (PCA) techniques to improve the quality of training data. Our approach achieves good results outperforming other popular models, which used the same principle of training with limited data. In addition, a detailed explanation of these techniques and some recommendations are provided. Our methodology constitutes a useful strategy for increasing samples in the deep learning domain and can be applied to improve segmentation accuracy. We believe that our strategy has a considerable interest in various applications such as medical and biological fields, especially in the early stages of experiments where there are few samples.
A DEEP LEARNING APPROACH FOR SEMANTIC SEGMENTATION IN BRAIN TUMOR IMAGESPNandaSai
Digital image processing is vast fields which can be using various applications. Which include Detection of criminal face, fingerprint authentication system, in medical field, object recognition etc. Brain tumor detection plays an important role in medical field. Brain tumor detection is detection of tumor affected part in the brain along with its shape size and boundary, so it useful in medical field.
Segmentation and the subsequent quantitative assessment of lesions in medical images provide valuable information for the analysis of neuropathologist and are important for planning of treatment strategies, monitoring of disease progression and prediction of patient outcome. For a better understanding of the pathophysiology of diseases, quantitative imaging can reveal clues about the disease characteristics and effects on particular anatomical structures
A new approachto image classification based on adeep multiclass AdaBoosting e...IJECEIAES
In recent years, deep learning methods have been developed in order to solve the problems. These methods were effective in solving complex problems. Convolution is one of the learning methods. This method is applied in classifying and processing of images as well. Hybrid methods are another multi-component machine learning method. These methods are categorized into independent and dependent types. Ada-Boosting algorithm is one of these methods. Today, the classification of images has many applications. So far, several algorithms have been presented for binary and multi-class classification. Most of the above-mentioned methods have a high dependence on the data. The present study intends to use a combination of deep learning methods and associated hybrid methods to classify the images. It is presumed that this method is able to reduce the error rate in images classification. The proposed algorithm consists of the Ada-Boosting hybrid method and bi-layer convolutional learning method. The proposed method was analyzed after it was implemented on a multi-class Mnist data set and displayed the result of the error rate reduction. The results of this study indicate that the error rate of the proposed method is less than Ada-Boosting and convolution methods. Also, the network has more stability compared to the other methods.
Validation Study of Dimensionality Reduction Impact on Breast Cancer Classifi...ijcsit
A fundamental problem in machine learning is identifying the most representative subset of features from
which we can construct a predictive model for a classification task. This paper aims to present a validation
study of dimensionality reduction effect on the classification accuracy of mammographic images. The
studied dimensionality reduction methods were: locality-preserving projection (LPP), locally linear
embedding (LLE), Isometric Mapping (ISOMAP) and spectral regression (SR). We have achieved high
rates of classifications. In some combinations the classification rate was 100%. But in most of the cases the
classification rate is about 95%. It was also found that the classification rate increases with the size of the
reduced space and the optimal value of space dimension is 60. We proceeded to validate the obtained
results by measuring some validation indices such as: Xie-Beni index, Dun index and Alternative Dun
index. The measurement of these indices confirms that the optimal value of reduced space dimension is
d=60.
Proposing a new method of image classification based on the AdaBoost deep bel...TELKOMNIKA JOURNAL
Image classification has different applications. Up to now, various algorithms have been presented
for image classification. Each of these methods has its own weaknesses and strengths. Reducing error rate
is an issue which many researches have been carried out about it. This research intends to optimize
the problem with hybrid methods and deep learning. The hybrid methods were developed to improve
the results of the single-component methods. On the other hand, a deep belief network (DBN) is a generative
probabilistic modelwith multiple layers of latent variables and is used to solve the unlabeled problems. In
fact, this method is anunsupervised method, in which all layers are one-way directed layers except for
the last layer. So far, various methods have been proposed for image classification, and the goal of this
research project was to use a combination of the AdaBoost method and the deep belief network method to
classify images. The other objective was to obtain better results than the previous results. In this project, a
combination of the deep belief network and AdaBoost method was used to boost learning and the network
potential was enhanced by making the entire network recursive. This method was tested on the MINIST
dataset and the results were indicative of a decrease in the error rate with the proposed method as compared
to the AdaBoost and deep belief network methods.
DEEP LEARNING FOR SMART GRID INTRUSION DETECTION: A HYBRID CNN-LSTM-BASED MODELgerogepatton
As digital technology becomes more deeply embedded in power systems, protecting the communication
networks of Smart Grids (SG) has emerged as a critical concern. Distributed Network Protocol 3 (DNP3)
represents a multi-tiered application layer protocol extensively utilized in Supervisory Control and Data
Acquisition (SCADA)-based smart grids to facilitate real-time data gathering and control functionalities.
Robust Intrusion Detection Systems (IDS) are necessary for early threat detection and mitigation because
of the interconnection of these networks, which makes them vulnerable to a variety of cyberattacks. To
solve this issue, this paper develops a hybrid Deep Learning (DL) model specifically designed for intrusion
detection in smart grids. The proposed approach is a combination of the Convolutional Neural Network
(CNN) and the Long-Short-Term Memory algorithms (LSTM). We employed a recent intrusion detection
dataset (DNP3), which focuses on unauthorized commands and Denial of Service (DoS) cyberattacks, to
train and test our model. The results of our experiments show that our CNN-LSTM method is much better
at finding smart grid intrusions than other deep learning algorithms used for classification. In addition,
our proposed approach improves accuracy, precision, recall, and F1 score, achieving a high detection
accuracy rate of 99.50%.
10th International Conference on Artificial Intelligence and Applications (AI...gerogepatton
10th International Conference on Artificial Intelligence and Applications (AI 2024) will provide an excellent international forum for sharing knowledge and results in theory, methodology and applications of Artificial Intelligence and its applications. The Conference looks for significant contributions to all major fields of the Artificial Intelligence, Soft Computing in theoretical and practical aspects. The aim of the Conference is to provide a platform to the researchers and practitioners from both academia as well as industry to meet and share cutting-edge development in the field.
International Journal of Artificial Intelligence & Applications (IJAIA)gerogepatton
The International Journal of Artificial Intelligence & Applications (IJAIA) is a bi monthly open access peer-reviewed journal that publishes articles which contribute new results in all areas of the Artificial Intelligence & Applications (IJAIA). It is an international journal intended for professionals and researchers in all fields of AI for researchers, programmers, and software and hardware manufacturers. The journal also aims to publish new attempts in the form of special issues on emerging areas in Artificial Intelligence and applications.
Immunizing Image Classifiers Against Localized Adversary Attacksgerogepatton
This paper addresses the vulnerability of deep learning models, particularly convolutional neural networks
(CNN)s, to adversarial attacks and presents a proactive training technique designed to counter them. We
introduce a novel volumization algorithm, which transforms 2D images into 3D volumetric representations.
When combined with 3D convolution and deep curriculum learning optimization (CLO), itsignificantly improves
the immunity of models against localized universal attacks by up to 40%. We evaluate our proposed approach
using contemporary CNN architectures and the modified Canadian Institute for Advanced Research (CIFAR-10
and CIFAR-100) and ImageNet Large Scale Visual Recognition Challenge (ILSVRC12) datasets, showcasing
accuracy improvements over previous techniques. The results indicate that the combination of the volumetric
input and curriculum learning holds significant promise for mitigating adversarial attacks without necessitating
adversary training.
May 2024 - Top 10 Read Articles in Artificial Intelligence and Applications (...gerogepatton
The International Journal of Artificial Intelligence & Applications (IJAIA) is a bi monthly open access peer-reviewed journal that publishes articles which contribute new results in all areas of the Artificial Intelligence & Applications (IJAIA). It is an international journal intended for professionals and researchers in all fields of AI for researchers, programmers, and software and hardware manufacturers. The journal also aims to publish new attempts in the form of special issues on emerging areas in Artificial Intelligence and applications.
3rd International Conference on Artificial Intelligence Advances (AIAD 2024)gerogepatton
3rd International Conference on Artificial Intelligence Advances (AIAD 2024) will act as a major forum for the presentation of innovative ideas, approaches, developments, and research projects in the area advanced Artificial Intelligence. It will also serve to facilitate the exchange of information between researchers and industry professionals to discuss the latest issues and advancement in the research area. Core areas of AI and advanced multi-disciplinary and its applications will be covered during the conferences.
International Journal of Artificial Intelligence & Applications (IJAIA)gerogepatton
The International Journal of Artificial Intelligence & Applications (IJAIA) is a bi monthly open access peer-reviewed journal that publishes articles which contribute new results in all areas of the Artificial Intelligence & Applications (IJAIA). It is an international journal intended for professionals and researchers in all fields of AI for researchers, programmers, and software and hardware manufacturers. The journal also aims to publish new attempts in the form of special issues on emerging areas in Artificial Intelligence and applications.
Information Extraction from Product Labels: A Machine Vision Approachgerogepatton
This research tackles the challenge of manual data extraction from product labels by employing a blend of
computer vision and Natural Language Processing (NLP). We introduce an enhanced model that combines
Convolutional Neural Networks (CNN) and Recurrent Neural Networks (RNN) in a Convolutional
Recurrent Neural Network (CRNN) for reliable text recognition. Our model is further refined by
incorporating the Tesseract OCR engine, enhancing its applicability in Optical Character Recognition
(OCR) tasks. The methodology is augmented by NLP techniques and extended through the Open Food
Facts API (Application Programming Interface) for database population and text-only label prediction.
The CRNN model is trained on encoded labels and evaluated for accuracy on a dedicated test set.
Importantly, our approach enables visually impaired individuals to access essential information on
product labels, such as directions and ingredients. Overall, the study highlights the efficacy of deep
learning and OCR in automating label extraction and recognition.
10th International Conference on Artificial Intelligence and Applications (AI...gerogepatton
10th International Conference on Artificial Intelligence and Applications (AI 2024) will provide an excellent international forum for sharing knowledge and results in theory, methodology and applications of Artificial Intelligence and its applications. The Conference looks for significant contributions to all major fields of the Artificial Intelligence, Soft Computing in theoretical and practical aspects. The aim of the Conference is to provide a platform to the researchers and practitioners from both academia as well as industry to meet and share cutting-edge development in the field.
International Journal of Artificial Intelligence & Applications (IJAIA)gerogepatton
The International Journal of Artificial Intelligence & Applications (IJAIA) is a bi monthly open access peer-reviewed journal that publishes articles which contribute new results in all areas of the Artificial Intelligence & Applications (IJAIA). It is an international journal intended for professionals and researchers in all fields of AI for researchers, programmers, and software and hardware manufacturers. The journal also aims to publish new attempts in the form of special issues on emerging areas in Artificial Intelligence and applications.
Research on Fuzzy C- Clustering Recursive Genetic Algorithm based on Cloud Co...gerogepatton
Aiming at the problems of poor local search ability and precocious convergence of fuzzy C-cluster
recursive genetic algorithm (FOLD++), a new fuzzy C-cluster recursive genetic algorithm based on
Bayesian function adaptation search (TS) was proposed by incorporating the idea of Bayesian function
adaptation search into fuzzy C-cluster recursive genetic algorithm. The new algorithm combines the
advantages of FOLD++ and TS. In the early stage of optimization, fuzzy C-cluster recursive genetic
algorithm is used to get a good initial value, and the individual extreme value pbest is put into Bayesian
function adaptation table. In the late stage of optimization, when the searching ability of fuzzy C-cluster
recursive genetic is weakened, the short term memory function of Bayesian function adaptation table in
Bayesian function adaptation search algorithm is utilized. Make it jump out of the local optimal solution,
and allow bad solutions to be accepted during the search. The improved algorithm is applied to function
optimization, and the simulation results show that the calculation accuracy and stability of the algorithm
are improved, and the effectiveness of the improved algorithm is verified
International Journal of Artificial Intelligence & Applications (IJAIA)gerogepatton
The International Journal of Artificial Intelligence & Applications (IJAIA) is a bi monthly open access peer-reviewed journal that publishes articles which contribute new results in all areas of the Artificial Intelligence & Applications (IJAIA). It is an international journal intended for professionals and researchers in all fields of AI for researchers, programmers, and software and hardware manufacturers. The journal also aims to publish new attempts in the form of special issues on emerging areas in Artificial Intelligence and applications.
10th International Conference on Artificial Intelligence and Soft Computing (...gerogepatton
10th International Conference on Artificial Intelligence and Soft Computing (AIS 2024) will
provide an excellent international forum for sharing knowledge and results in theory, methodology, and
applications of Artificial Intelligence, Soft Computing. The Conference looks for significant
contributions to all major fields of the Artificial Intelligence, Soft Computing in theoretical and practical
aspects. The aim of the Conference is to provide a platform to the researchers and practitioners from
both academia as well as industry to meet and share cutting-edge development in the field.
International Journal of Artificial Intelligence & Applications (IJAIA)gerogepatton
Employee attrition refers to the decrease in staff numbers within an organization due to various reasons.
As it has a negative impact on long-term growth objectives and workplace productivity, firms have
recognized it as a significant concern. To address this issue, organizations are increasingly turning to
machine-learning approaches to forecast employee attrition rates. This topic has gained significant
attention from researchers, especially in recent times. Several studies have applied various machinelearning methods to predict employee attrition, producing different resultsdepending on the employed
methods, factors, and datasets. However, there has been no comprehensive comparative review of multiple
studies applying machine-learning models to predict employee attrition to date. Therefore, this study aims
to fill this gap by providing an overview of research conducted on applying machine learning to predict
employee attrition from 2019 to February 2024. A literature review of relevant studies was conducted,
summarized, and classified. Most studies agree on conducting comparative experiments with multiple
predictive models to determine the most effective one.From this literature survey, the RF algorithm and
XGB ensemble method are repeatedly the best-performing, outperforming many other algorithms.
Additionally, the application of deep learning to employee attrition prediction issues also shows promise.
While there are discrepancies in the datasets used in previous studies, it is notable that the dataset
provided by IBM is the most widely utilized. This study serves as a concise review for new researchers,
facilitating their understanding of the primary techniques employed in predicting employee attrition and
highlighting recent research trends in this field. Furthermore, it provides organizations with insight into
the prominent factors affecting employee attrition, as identified by studies, enabling them to implement
solutions aimed at reducing attrition rates.
10th International Conference on Artificial Intelligence and Applications (AI...gerogepatton
10th International Conference on Artificial Intelligence and Applications (AIFU 2024) is a forum for presenting new advances and research results in the fields of Artificial Intelligence. The conference will bring together leading researchers, engineers and scientists in the domain of interest from around the world. The scope of the conference covers all theoretical and practical aspects of the Artificial Intelligence.
International Journal of Artificial Intelligence & Applications (IJAIA)gerogepatton
The International Journal of Artificial Intelligence & Applications (IJAIA) is a bi monthly open access peer-reviewed journal that publishes articles which contribute new results in all areas of the Artificial Intelligence & Applications (IJAIA). It is an international journal intended for professionals and researchers in all fields of AI for researchers, programmers, and software and hardware manufacturers. The journal also aims to publish new attempts in the form of special issues on emerging areas in Artificial Intelligence and applications.
THE TRANSFORMATION RISK-BENEFIT MODEL OF ARTIFICIAL INTELLIGENCE:BALANCING RI...gerogepatton
This paper summarizes the most cogent advantages and risks associated with Artificial Intelligence from an
in-depth review of the literature. Then the authors synthesize the salient risk-related models currently being
used in AI, technology and business-related scenarios. Next, in view of an updated context of AI along with
theories and models reviewed and expanded constructs, the writers propose a new framework called “The
Transformation Risk-Benefit Model of Artificial Intelligence” to address the increasing fears and levels of
AIrisk. Using the model characteristics, the article emphasizes practical and innovative solutions where
benefitsoutweigh risks and three use cases in healthcare, climate change/environment and cyber security to
illustrate unique interplay of principles, dimensions and processes of this powerful AI transformational
model.
13th International Conference on Software Engineering and Applications (SEA 2...gerogepatton
13th International Conference on Software Engineering and Applications (SEA 2024) will provide an excellent international forum for sharing knowledge and results in theory, methodology and applications of Software Engineering and Applications. The goal of this conference is to bring together researchers and practitioners from academia and industry to focus on understanding Modern software engineering concepts and establishing new collaborations in these areas.
International Journal of Artificial Intelligence & Applications (IJAIA)gerogepatton
The International Journal of Artificial Intelligence & Applications (IJAIA) is a bi monthly open access peer-reviewed journal that publishes articles which contribute new results in all areas of the Artificial Intelligence & Applications (IJAIA). It is an international journal intended for professionals and researchers in all fields of AI for researchers, programmers, and software and hardware manufacturers. The journal also aims to publish new attempts in the form of special issues on emerging areas in Artificial Intelligence and applications.
AN IMPROVED MT5 MODEL FOR CHINESE TEXT SUMMARY GENERATIONgerogepatton
Complicated policy texts require a lot of effort to read, so there is a need for intelligent interpretation of
Chinese policies. To better solve the Chinese Text Summarization task, this paper utilized the mT5 model
as the core framework and initial weights. Additionally, In addition, this paper reduced the model size
through parameter clipping, used the Gap Sentence Generation (GSG) method as unsupervised method,
and improved the Chinese tokenizer. After training on a meticulously processed 30GB Chinese training
corpus, the paper developed the enhanced mT5-GSG model. Then, when fine-tuning the Chinese Policy
text, this paper chose the idea of “Dropout Twice”, and innovatively combined the probability distribution
of the two Dropouts through the Wasserstein distance. Experimental results indicate that the proposed
model achieved Rouge-1, Rouge-2, and Rouge-L scores of 56.13%, 45.76%, and 56.41% respectively on
the Chinese policy text summarization dataset.
Connector Corner: Automate dynamic content and events by pushing a buttonDianaGray10
Here is something new! In our next Connector Corner webinar, we will demonstrate how you can use a single workflow to:
Create a campaign using Mailchimp with merge tags/fields
Send an interactive Slack channel message (using buttons)
Have the message received by managers and peers along with a test email for review
But there’s more:
In a second workflow supporting the same use case, you’ll see:
Your campaign sent to target colleagues for approval
If the “Approve” button is clicked, a Jira/Zendesk ticket is created for the marketing design team
But—if the “Reject” button is pushed, colleagues will be alerted via Slack message
Join us to learn more about this new, human-in-the-loop capability, brought to you by Integration Service connectors.
And...
Speakers:
Akshay Agnihotri, Product Manager
Charlie Greenberg, Host
Transcript: Selling digital books in 2024: Insights from industry leaders - T...BookNet Canada
The publishing industry has been selling digital audiobooks and ebooks for over a decade and has found its groove. What’s changed? What has stayed the same? Where do we go from here? Join a group of leading sales peers from across the industry for a conversation about the lessons learned since the popularization of digital books, best practices, digital book supply chain management, and more.
Link to video recording: https://bnctechforum.ca/sessions/selling-digital-books-in-2024-insights-from-industry-leaders/
Presented by BookNet Canada on May 28, 2024, with support from the Department of Canadian Heritage.
Search and Society: Reimagining Information Access for Radical FuturesBhaskar Mitra
The field of Information retrieval (IR) is currently undergoing a transformative shift, at least partly due to the emerging applications of generative AI to information access. In this talk, we will deliberate on the sociotechnical implications of generative AI for information access. We will argue that there is both a critical necessity and an exciting opportunity for the IR community to re-center our research agendas on societal needs while dismantling the artificial separation between the work on fairness, accountability, transparency, and ethics in IR and the rest of IR research. Instead of adopting a reactionary strategy of trying to mitigate potential social harms from emerging technologies, the community should aim to proactively set the research agenda for the kinds of systems we should build inspired by diverse explicitly stated sociotechnical imaginaries. The sociotechnical imaginaries that underpin the design and development of information access technologies needs to be explicitly articulated, and we need to develop theories of change in context of these diverse perspectives. Our guiding future imaginaries must be informed by other academic fields, such as democratic theory and critical theory, and should be co-developed with social science scholars, legal scholars, civil rights and social justice activists, and artists, among others.
The Art of the Pitch: WordPress Relationships and SalesLaura Byrne
Clients don’t know what they don’t know. What web solutions are right for them? How does WordPress come into the picture? How do you make sure you understand scope and timeline? What do you do if sometime changes?
All these questions and more will be explored as we talk about matching clients’ needs with what your agency offers without pulling teeth or pulling your hair out. Practical tips, and strategies for successful relationship building that leads to closing the deal.
GDG Cloud Southlake #33: Boule & Rebala: Effective AppSec in SDLC using Deplo...James Anderson
Effective Application Security in Software Delivery lifecycle using Deployment Firewall and DBOM
The modern software delivery process (or the CI/CD process) includes many tools, distributed teams, open-source code, and cloud platforms. Constant focus on speed to release software to market, along with the traditional slow and manual security checks has caused gaps in continuous security as an important piece in the software supply chain. Today organizations feel more susceptible to external and internal cyber threats due to the vast attack surface in their applications supply chain and the lack of end-to-end governance and risk management.
The software team must secure its software delivery process to avoid vulnerability and security breaches. This needs to be achieved with existing tool chains and without extensive rework of the delivery processes. This talk will present strategies and techniques for providing visibility into the true risk of the existing vulnerabilities, preventing the introduction of security issues in the software, resolving vulnerabilities in production environments quickly, and capturing the deployment bill of materials (DBOM).
Speakers:
Bob Boule
Robert Boule is a technology enthusiast with PASSION for technology and making things work along with a knack for helping others understand how things work. He comes with around 20 years of solution engineering experience in application security, software continuous delivery, and SaaS platforms. He is known for his dynamic presentations in CI/CD and application security integrated in software delivery lifecycle.
Gopinath Rebala
Gopinath Rebala is the CTO of OpsMx, where he has overall responsibility for the machine learning and data processing architectures for Secure Software Delivery. Gopi also has a strong connection with our customers, leading design and architecture for strategic implementations. Gopi is a frequent speaker and well-known leader in continuous delivery and integrating security into software delivery.
Dev Dives: Train smarter, not harder – active learning and UiPath LLMs for do...UiPathCommunity
💥 Speed, accuracy, and scaling – discover the superpowers of GenAI in action with UiPath Document Understanding and Communications Mining™:
See how to accelerate model training and optimize model performance with active learning
Learn about the latest enhancements to out-of-the-box document processing – with little to no training required
Get an exclusive demo of the new family of UiPath LLMs – GenAI models specialized for processing different types of documents and messages
This is a hands-on session specifically designed for automation developers and AI enthusiasts seeking to enhance their knowledge in leveraging the latest intelligent document processing capabilities offered by UiPath.
Speakers:
👨🏫 Andras Palfi, Senior Product Manager, UiPath
👩🏫 Lenka Dulovicova, Product Program Manager, UiPath
Neuro-symbolic is not enough, we need neuro-*semantic*Frank van Harmelen
Neuro-symbolic (NeSy) AI is on the rise. However, simply machine learning on just any symbolic structure is not sufficient to really harvest the gains of NeSy. These will only be gained when the symbolic structures have an actual semantics. I give an operational definition of semantics as “predictable inference”.
All of this illustrated with link prediction over knowledge graphs, but the argument is general.
DevOps and Testing slides at DASA ConnectKari Kakkonen
My and Rik Marselis slides at 30.5.2024 DASA Connect conference. We discuss about what is testing, then what is agile testing and finally what is Testing in DevOps. Finally we had lovely workshop with the participants trying to find out different ways to think about quality and testing in different parts of the DevOps infinity loop.
JMeter webinar - integration with InfluxDB and GrafanaRTTS
Watch this recorded webinar about real-time monitoring of application performance. See how to integrate Apache JMeter, the open-source leader in performance testing, with InfluxDB, the open-source time-series database, and Grafana, the open-source analytics and visualization application.
In this webinar, we will review the benefits of leveraging InfluxDB and Grafana when executing load tests and demonstrate how these tools are used to visualize performance metrics.
Length: 30 minutes
Session Overview
-------------------------------------------
During this webinar, we will cover the following topics while demonstrating the integrations of JMeter, InfluxDB and Grafana:
- What out-of-the-box solutions are available for real-time monitoring JMeter tests?
- What are the benefits of integrating InfluxDB and Grafana into the load testing stack?
- Which features are provided by Grafana?
- Demonstration of InfluxDB and Grafana using a practice web application
To view the webinar recording, go to:
https://www.rttsweb.com/jmeter-integration-webinar
Builder.ai Founder Sachin Dev Duggal's Strategic Approach to Create an Innova...Ramesh Iyer
In today's fast-changing business world, Companies that adapt and embrace new ideas often need help to keep up with the competition. However, fostering a culture of innovation takes much work. It takes vision, leadership and willingness to take risks in the right proportion. Sachin Dev Duggal, co-founder of Builder.ai, has perfected the art of this balance, creating a company culture where creativity and growth are nurtured at each stage.
Slack (or Teams) Automation for Bonterra Impact Management (fka Social Soluti...Jeffrey Haguewood
Sidekick Solutions uses Bonterra Impact Management (fka Social Solutions Apricot) and automation solutions to integrate data for business workflows.
We believe integration and automation are essential to user experience and the promise of efficient work through technology. Automation is the critical ingredient to realizing that full vision. We develop integration products and services for Bonterra Case Management software to support the deployment of automations for a variety of use cases.
This video focuses on the notifications, alerts, and approval requests using Slack for Bonterra Impact Management. The solutions covered in this webinar can also be deployed for Microsoft Teams.
Interested in deploying notification automations for Bonterra Impact Management? Contact us at sales@sidekicksolutionsllc.com to discuss next steps.
Accelerate your Kubernetes clusters with Varnish CachingThijs Feryn
A presentation about the usage and availability of Varnish on Kubernetes. This talk explores the capabilities of Varnish caching and shows how to use the Varnish Helm chart to deploy it to Kubernetes.
This presentation was delivered at K8SUG Singapore. See https://feryn.eu/presentations/accelerate-your-kubernetes-clusters-with-varnish-caching-k8sug-singapore-28-2024 for more details.
"Impact of front-end architecture on development cost", Viktor TurskyiFwdays
I have heard many times that architecture is not important for the front-end. Also, many times I have seen how developers implement features on the front-end just following the standard rules for a framework and think that this is enough to successfully launch the project, and then the project fails. How to prevent this and what approach to choose? I have launched dozens of complex projects and during the talk we will analyze which approaches have worked for me and which have not.
State of ICS and IoT Cyber Threat Landscape Report 2024 previewPrayukth K V
The IoT and OT threat landscape report has been prepared by the Threat Research Team at Sectrio using data from Sectrio, cyber threat intelligence farming facilities spread across over 85 cities around the world. In addition, Sectrio also runs AI-based advanced threat and payload engagement facilities that serve as sinks to attract and engage sophisticated threat actors, and newer malware including new variants and latent threats that are at an earlier stage of development.
The latest edition of the OT/ICS and IoT security Threat Landscape Report 2024 also covers:
State of global ICS asset and network exposure
Sectoral targets and attacks as well as the cost of ransom
Global APT activity, AI usage, actor and tactic profiles, and implications
Rise in volumes of AI-powered cyberattacks
Major cyber events in 2024
Malware and malicious payload trends
Cyberattack types and targets
Vulnerability exploit attempts on CVEs
Attacks on counties – USA
Expansion of bot farms – how, where, and why
In-depth analysis of the cyber threat landscape across North America, South America, Europe, APAC, and the Middle East
Why are attacks on smart factories rising?
Cyber risk predictions
Axis of attacks – Europe
Systemic attacks in the Middle East
Download the full report from here:
https://sectrio.com/resources/ot-threat-landscape-reports/sectrio-releases-ot-ics-and-iot-security-threat-landscape-report-2024/
State of ICS and IoT Cyber Threat Landscape Report 2024 preview
A Modified CNN-Based Face Recognition System
1. International Journal of Artificial Intelligence and Applications (IJAIA), Vol.12, No.2, March 2021
DOI: 10.5121/ijaia.2021.12201 1
A MODIFIED CNN-BASED FACE
RECOGNITION SYSTEM
Jayanthi Raghavan and Majid Ahmadi
Department of Electrical and Computer Engineering,
University of Windsor, Windsor, Canada
ABSTRACT
In this work, deep CNN based model have been suggested for face recognition. CNN is employed to extract
unique facial features and softmax classifier is applied to classify facial images in a fully connected layer
of CNN. The experiments conducted in Extended YALE B and FERET databases for smaller batch sizes
and low value of learning rate, showed that the proposed model has improved the face recognition
accuracy. Accuracy rates of up to 96.2% is achieved using the proposed model in Extended Yale B
database. To improve the accuracy rate further, preprocessing techniques like SQI, HE, LTISN, GIC and
DoG are applied to the CNN model. After the application of preprocessing techniques, the improved
accuracy of 99.8% is achieved with deep CNN model for the YALE B Extended Database. In FERET
Database with frontal face, before the application of preprocessing techniques, CNN model yields the
maximum accuracy of 71.4%. After applying the above-mentioned preprocessing techniques, the accuracy
is improved to 76.3%
KEYWORDS
CNN, ANN, GPU
1. INTRODUCTION
For human brain, recognizing face is a very simple task and can be performed fast. In computer
vision, face recognition is a very challenging task. Even though the face recognition research is in
an advanced state [7], till now it is not possible to obtain results on par with humans.
To date, many approaches have been suggested for facial recognition. Holistic method works by
projecting facial images onto a low-dimensional space, which neglects surplus details and
variations that are not needed for the facial recognition [11]. One of the methods under this
category is PCA [2]. The holistic methods are sensitive to local distortions like facial expression
or illumination variation.
Subsequently, progress in the field of computer vision led to the growth of feature-based method
in which features are extracted from various parts of a face image. Feature-based methods are
robust to local variations such as intensity variation and face expression changes [22]. With the
development of the local feature descriptors, feature based methods gained popularity. Local
Binary Pattern (LBP) [6] is an extensively applied local feature descriptor in face recognition.
The recent trend is towards neural network-based approach [30]. Deep learning-based methods
achieve excellent results in many fields like robotics and autonomous driving cars [3]. Deep
learning methods are based on convolutional neural networks (CNNs). CNNs are slightlydifferent
from normal neural network. In CNN, neurons in convolutional layer are thinly connected to the
2. International Journal of Artificial Intelligence and Applications (IJAIA), Vol.12, No.2, March 2021
2
neurons in the next layer based on their relative location. CNNs are multilayer network trained
from end to end with raw image pixel values assigned to classifier outputs.
The main advantage of deep learning methods is that they can be trained with very big datasets to
learn the vital features to represent the input data. The main issue with deep learning method is
models trained with small datasets are having the problem of poor generalization, which results in
over-fitting.
Generalization term indicates the performance difference of a network model when assessed on
earlier viewed training data against the testing data, the network has never viewed before [8].
Models with poor generalizability have overfitted the training data.
Overfitting is a term used when the network model functions extremely good with the training
data, but could not work well with the test data. In the overfitted network validation error goes up
while the training error comes down [21].
To reduce the overfitting, regularization process is employed on the network model.
Regularization is a method of making minute changes to the actual network model and the
learning algorithm, so that the model functions better in both training and testing data set.
Regularization is defined as “Allowing to generalize well to unseen data even when training on a
finite training set or with an imperfect optimization procedure” [4]. There are various
regularization techniques available in machine learning like Dropout, Data Augmentation, Early
stopping, batch normalization etc. [10]. Some of the regularization techniques are explained
below.
Dropout means removing units temporarily in a neural network, together with all the incoming
and outgoing links [23]. Dropout can be explained as the regularization technique by including
noise to the hidden units of the network.
Another popular technique is batch normalization. Batch Normalization operates by deducting
the batch mean from each activation and dividing by the standard deviation of the batch [8]. The
normalization technique together with standardization is used as a typical combination in the
preprocessing of pixel values. Batch normalization technique can be employed to any individual
layer within the network. Hence it is powerful [5].
Data augmentation is an extensively applied technique in deep learning. The performance of the
deep neural network is based on the size of the dataset [45]. Deep learning network needs large
dataset to avoid the problem of overfitting. Sometimes it is difficult to get quality and huge
database especially in medical field. Data augmentation helps to artificially inflate the size of the
training dataset by the methods called data warping or oversampling. The augmented data is a
representation of detailed set of feasible data points, thus minimizing the distance between the
training and the validation set.
Data augmentation techniques are useful to provide powerful regularization in terms of improved
generalization which in turn yields better network performance [46]. Bengio et al [48]
demonstrate that the data augmentation techniques are very effective in deep networks compared
to shallow networks. In addition to flipping and cropping various techniques like color casting,
vignetting, rotation, horizontal and vertical stretching proven to be effective data augmentation
methods [49].
The main aim of augmentation technique is to diminish the effect of overfitting on models using
traditional transformations to manipulate the training data. As an example, a labeled image in the
3. International Journal of Artificial Intelligence and Applications (IJAIA), Vol.12, No.2, March 2021
3
dataset can be increased by various processes like flipping, rotation, morphing and zooming.
After some time, trained network gets exposure to such modifications and the it can identify the
same object with different variations.
There are some limitations associated with the application of data augmentation techniques. One
of the main disadvantages is it demands extra memory, computational expenses involved with
implementing augmentation methods and longer training time. Some geometric transformations
such as translation or random cropping must be monitored carefully to ensure that the labels of
the images are not altered during the transformation process [8].
Optimizers update the weight parameters to minimize the cost function. Cost function is defined
as the difference between predicted and actual output. One of the popular optimizers is Adam
optimizer [1].
Adam is an adaptive learning rate optimization algorithm, which calculates learning rates for
different parameters individually. Adam uses computations of first and second moments of
gradient to adjust the learning rate for each weight of the network.
Application of regularization methods help to improve the accuracy rates. To increase the
accuracy rates further, preprocessing techniques are applied to deep CNN architecture.
2. PREPROCESSING METHODS
Face recognition task becomes challenging due to illumination conditions, occlusion, pose and
facial expression variations [33]. The variation in illumination is one of the main challenging
problems which affects the performance of the face recognition system. Among all, shadowing
effect, underexposure, and overexposure conditions are challenging problems that need to be
addressed in the face recognition process [35]. If the lighting conditions present in the gallery
image is different from the probe image, then the process of face recognition may completely fail
[34]. A good face recognition system should be able to give accurate recognition rate under the
different illumination conditions between images of the same face [32]. Image enhancement
algorithms play a great role in handling the illumination variation.
The main aim of preprocessing is to remove features that obstruct the process of classifying the
images of the same person (within-class differences), thereby boosting the difference of them
with others (between-class differences) [36].
For better face recognition under uncontrolled and illumination variation conditions, the vital
features responsible for differentiating two different faces require to be retained. The shadows
produced in facial images due to variation in lighting directions may cause loss of important
facial features which are helpful for recognition. A preprocessing method must enhance the
intensity in the regions of inadequately illuminated and decrease the intensity in the densely
illuminated regions while retaining the intensity in the fairly illuminated portions [37]. Few
important preprocessing techniques are discussed below.
2.1. Histogram Equalization (HE)
HE [38] flattens the histogram of the image and expands the dynamic range of the pixel intensity
values by employing cumulative density function. The Histogram Equalization is a global
preprocessing technique.
4. International Journal of Artificial Intelligence and Applications (IJAIA), Vol.12, No.2, March 2021
4
For image I(x, y) with discrete k. gray values histogram is defined by the probability of
occurrence of the gray level I [39] is given by equation (1) as follows.
p(i) = 𝑛𝑖 /N (1)
Where i∈ 0, 1…k -1 grey level and N is total number of pixels in the image.
2.2. Self Quotient Image (SQI)
SQI [40] is one of the illumination invariant algorithm suggested for handling both shadow and
lighting changes. It is defined as the ratio of the intensity of the input image to its smooth version
as given in equations (2) and (3).
Q(x,y)=I(x,y)/S(x,y) (2)
= I(x,y) /(F(x,y) *I(x,y)) (3)
where I (x, y) is the face image and S (x, y) is a smoothed version of the image and ∗ is the
convolution operation. F is the smoothing kernel which in this case is a weighted Gaussian filter
and Q is the Self Quotient Image since it is derived from one image and has the same quotient
form as that in the quotient image method.
2.3. Locally Tuned Inverse Sine Nonlinear (LTISN)
LTISN [41] is a nonlinear and pixel by pixel approach, where the improved intensity values are
calculated by applying the inverse sine function with a tunable parameter based on the nearby
pixel values given in the equations (4), (5), (6), (7) and (8). The intensity range of the image is
rescaled to [0 1] followed by a nonlinear transfer function.
2.4.Gamma Intensity Correction (GIC)
GIC [42] is a nonlinear gray-level transformation that substitutes gray-level I with the gray level
𝐼
1
ϒ, given by the equation (9).
𝐼 = 𝐼
1
ϒ(9)
5. International Journal of Artificial Intelligence and Applications (IJAIA), Vol.12, No.2, March 2021
5
As shown in the (Fig.1), for the values of gamma less than 1.0 darkens the image and for the
values of gamma greater than 1.0 lightens the image. When gamma value is 1.0, it does not
produce any effect.
Fig.1. Gamma Intensity Correction
2.5. Difference of Gaussian (DoG)
DoG is a grayscale image enhancement algorithm [42] that involves the subtraction of one
blurred version of an original grayscale image from another less blurred version of the original.
The blurred images are obtained by convolving the original grayscale image with Gaussian
kernels having differing standard deviations [34], which is given in the equation (10).
𝜎1, 𝜎2 are Gaussian kernel widths.
2.6. Contrast-Limited Adaptive Histogram Equalization (CLAHE)
CLAHE works on small areas in the image, known as tiles, rather than the whole image [43].
Individual tile's contrast is improved. Hence histogram of the output area is roughly matching
with the histogram specified by the distribution parameter. The tiles present in the neighborhood
regions are then joined by applying bilinear interpolation to minimize the effect of artificially
induced border line [39]. In CLAHE, the image is divided into a limited number of regions and
the same histogram equalization technique is applied to pixels in each region [44].
3. CONVOLUTIONAL NEURAL NETWORK
Deep learning-based methods have shown better performances in terms of accuracy and speed of
processing in image recognition
3.1. CNN Basics
CNN is biologically inspired by visual cortex in brain [20]. Different phases of learning
procedure in CNN is similar to the visual cortex. The visual cortex has small group of cells that
are reactive to particular regions of the visual field. Hubel and Wiesel [17] experimentally
showed that particular group of neurons in the cat’s brain reacted to the appearance of edges of a
certain orientation as shown in the (Fig.2). Further they illustrated that one specific set of neurons
6. International Journal of Artificial Intelligence and Applications (IJAIA), Vol.12, No.2, March 2021
6
were activated, when exhibited to vertical edges and also another set of neurons fired when
exposed to horizontal or diagonal edges. All of these categories of neurons are arranged in a
columnar configuration and collectively these neurons can produce visual perception.
Different portions of the visual cortex are classified as V1, V2, V3, and V4. In general, V1 and
V2 regions of visual cortex are having close resemblance with convolutional and subsampling
layers, whereas inferior temporal region resembles the higher layers of CNN [16]. During
theprocess of training, CNN learns with the help of backpropagation algorithm by making
adjustments in weights with respect to the target.
Fig.2. Hubel and Wiesel experiment
3.2. CNN Architecture
In CNN architecture, network layers are divided into three types: the convolutional, pooling and
fully connected layers [20]. The architecture of CNN is as shown in Fig 3.
3.2.1. Convolutional Layer
In CNN, every neuron in the convolutional layer is linked only to a small portion of the neurons
in the preceding layer, which is in square shape area across the height and width dimensions. The
size of this square is a hyperparameter (controllable parameter) and called as Receptive Field. For
the depth dimension, there is no hyperparameter available since the convolution operations are
normally carried out for the whole depth. Generally, the depth dimension of the input describes
the various colors of the image. Hence it is usually required to link them in order to bring out
necessary information.
Fig.3.The CNN architecture
7. International Journal of Artificial Intelligence and Applications (IJAIA), Vol.12, No.2, March 2021
7
Neurons present in the convolution operator can recognize certain local patterns of the previous
layer’s output. Once features are obtained, its actual position is not important and all the neurons
are expected to identify the same pattern. This is achieved by forcing all the neurons to have a
common single set of parameters known as Parameter Sharing [9].
In order to identify various unique features within one layer, it is necessary to have multiple
filters, where each filter is a group of neurons that recognize a certain pattern at different
locations in the image.
3.2.2. Pooling Layer
The main aim of pooling layer is to reduce the complexity of CNNs. The neurons present in the
pooling layer form a square shaped area across the width and height dimensions of the preceding
layer. Even though it is very similar to the convolutional layer, it is different from convolutional
layer because the pooling layer is Non-Parametrized Layer.
The function carried out by this layer is known as subsampling or down sampling. During this
process, contraction in size results in concurrent loss of data. On the other hand, such a loss is
helpful to the network because the reduction in size not only reduces the computational burden
for the succeeding layers of the network and also it reduces the effects of overfitting [12].
Max pooling and average pooling are the generally used techniques shown in (Fig.4). Max
Pooling chooses the largest element within each receptive field [14] whereas Average Pooling
computes the average among the output neurons within the pooling window. Max-pooling
chooses the most prominent feature in a pooling window. On the other hand, average-pooling
method selects whole features into consideration. Thus, max-pooling method keeps texture
related information, while average pooling method retains the background related data [24].
Pooling operation does not combine neurons with different depth values. Instead, the resulting
pooling layer will have the uniform depth as the previous layer and it will only combine local
areas within a filter.
Fig.4.Max Pooling and Average Pooling Operations Illustration.
3.2.3. Fully Connected Layer
The filters and neurons present in this layer are connected to all the activation in the preceding
layers resulting in a completely connected structure. Hence the name. The output feature maps of
the final convolution or pooling layer is converted into a one-dimensional (1D) array of numbers
[18]. High-level reasoning in the network is carried out via fully connected layers [19].
8. International Journal of Artificial Intelligence and Applications (IJAIA), Vol.12, No.2, March 2021
8
The final fully connected layer has the same number of output nodes as the number of classes.
Each fully connected layer is followed by a nonlinear function, such as ReLU (Rectified Linear
Units). ReLU is an activation function operates by thresholding values at 0, i.e. f (x) = max (0, x).
In other words, it outputs 0 when x < 0, and contrarily, it outputs a linear function with a slope of
1 when x ≥ 0 [15] as shown in the (Fig.5).
Fig.5. The Rectified Linear Unit (ReLU)
3.3. CNN Operation
Based on local receptive field, each component in a convolutional layer accepts inputs from a set
of adjacent units belonging to the preceding layer layer. This way neurons are proficient in
extracting rudimentary features like edges or corners. These features are then linked by the
succeeding convolutional layers in order to further extract high level features.
The components of a convolutional layer are arranged in planes. All units of a plane share the
same set of weights. Thus, each plane is in charge for building a particular feature. The results
obtained from the plane is termed as feature maps. Each convolutional layer consists of several
planes, so that multiple feature maps can be constructed at each location. The most significant
features derived are passed from initial layers to higher layers. As the features are passed to the
higher layer, there is a dimensionality reduction in features determined by kernel size of the
convolutional and max-pooling layers. On the other hand, there is an increase in number of
feature maps for representing better features of the input images for ensuring classification
accuracy [13]. The derived feature vector either could be an input for classification task or could
be treated as a feature vector for next level processing.
The network designed to do the classification task consists of several convolution pooling layers
followed by some fully-connected layers. The first two layers mentioned above perform
convolution and pooling operation in order to extract high-level features. The final layer’s output
of CNN is applied as the input to a fully connected network, which does the task of classification.
The output layer for classification task consists of one neuron for each class and the values of
these neurons describes the score of each class. If score distribution is chosen, the score range
will be between zero and one. The summation of class scores is one. The values of each neuron
can be presumed as the probability of occurrence of the class.
9. International Journal of Artificial Intelligence and Applications (IJAIA), Vol.12, No.2, March 2021
9
Fig.6. A visual representation of the various hyperparameters of convolutional layers: receptive field,
stride and padding
4. EXPERIMENTAL SETUP
The implementations are carried out in MATLAB 2019b on a workstation with windows 10 OS,
AMD processor with 2.0 GHz, hard drive with 8 GB RAM. The experiments are conducted in
Extended Yale B and FERET Database with deep CNN model.
4.1. Extended Yale B Database
The extended Yale Face Database B contains 16128 images of 28 human subjects under 9 poses
and 64 illumination conditions as shown in (Fig.7). The data format of this database is the same
as that of the Yale Face Database B.
Fig.7. Images from extended Yale B database.
4.2. FERET Database
The FERET database was collected in 15 sessions between August 1993 and July 1996. The
database contains 1564 sets of images for a total of 14,126 images that includes 1199 individuals
and 365 duplicate sets of images as shown in (Fig.8). A duplicate set is a second set of images of
a person already in the database and was usually taken on a different day.
10. International Journal of Artificial Intelligence and Applications (IJAIA), Vol.12, No.2, March 2021
10
Fig.8. Images from FERET database
4.3. Experiment
The model uses a deep CNN network with six convolution layers (with size 3x3), two
maxpooling and two fully convolutional layers. The first two convolutions use 8 filters, next two
uses 16 and the last convolution layers uses 32 each in order to extract complex features.
The model uses batch normalization and dropouts in all convolutional layers along with ReLU
activation function. In order to reduce the overfitting, the dropout rate is kept higher on the final
convolutional layers. Two maxpooling layers are introduced after second and fourth
convolutional layers to reduce the feature space dimension. The two fully connected layers use
128 and 28 neurons respectively. The final classification layer uses softmax activation with
categorical cross entropy loss function. The softmax layer is used to produce the classification
scores, in which each score is the probability of a particular class for a given instance [31].
The dropout principle is employed on the convolutional neural networks model and the value of
drop out is chosen by trial and error method. The (Fig.9) shows the network model with drop out
and the values of the drop out is different for different layers. The model also uses Adam
Optimizer with a learning rate of 0.001 which was found empirically after trying different
combinations.
The feature maps are subsampled with maxpooling layers with a stride of 2x2. Stride is the
number of pixels which shifts over the input matrix. When the stride is 2, it means the filter is
moved by 2 pixels at a time. The number of neurons in this output layer is limited to 28 in
Extended Yale B database. In FERET database, the number of neurons in the output layer is 994.
Data Augmentation is implemented in the form of simple methods like resizing, rotation, and
reflection. The batch size of 128 and number of epochs 250 are chosen to achieve best results.
The training is designed to use different batch sizes to do a fair trade off between accuracy and
training time.
(a) (b)
Fig.9. (a) The neural networks model (b) The model after applying drop out
11. International Journal of Artificial Intelligence and Applications (IJAIA), Vol.12, No.2, March 2021
11
5. RESULTS
The CNN trained with back-propagation algorithm in batch mode with a various batch sizes of 4,
8, 16 and 32 for different epochs are computed for Extended YALE B with an image size of
128×128 for 50 epochs. The same experiment is conducted on FERET database for the frontal
face image size of 128×128 for 300 epochs.
For the Extended YALE B Database, the maximum accuracy rate of 97.2% is achieved for the
batch size of 4 without applying preprocessing technique as shown in (Fig.10). After applying the
preprocessing techniques, the maximum accuracy is improved to 99.8% as shown in (Fig.11).
For the batch size of 8, the maximum accuracy rate of 97.0% is achieved without applying any
preprocessing technique. After applying the preprocessing techniques, the maximum accuracy is
improved to 99.4%.
For the batch size of 8, SQI, LTISN and HE perform very equally well and yield the maximum
accuracy of 99.4%, 99.3% and 99.1%.
For the batch size of 16, the maximum accuracy rate of 96.8% is achieved without applying
preprocessing technique. After applying the preprocessing techniques, the maximum accuracy is
improved to 99.1%.
For the batch size of 16 HE, SQI and GIC perform equally well and yield the maximum accuracy
of 99.1%. 98.8% and 98.8% respectively.
For the batch size of 32, the maximum accuracy rate of 96.2% is achieved without applying
preprocessing technique. After applying the preprocessing techniques, the maximum accuracy is
improved to 98.7% for the same batch size.
For the batch size of 32, GIC, SQI, and HE perform equally well and yield the maximum
accuracy of 98.7%, 98.3% and 98.3% respectively. Table 1 shows the accuracy rates of Extended
Yale B database for various batch sizes before and after application of preprocessing techniques.
For the FERET Database, the maximum accuracy rate of 71.4% is achieved without applying
preprocessing technique, for the batch size of 4, as shown in (Fig.12). After applying the
preprocessing techniques, the maximum accuracy improved to 76.6% as shown in (Fig.13).
DoG, HE and LTISN perform equally well and yield the maximum accuracy of 76.6%, 76.3%
and 75.6% respectively.
For the batch size of 8, the maximum accuracy rate of 71.2% is achieved without applying
preprocessing technique. After applying the preprocessing techniques, the maximum accuracy is
improved to 76.4%.
For the batch size of 8, HE, SQI and LTISN perform very equally well and yield the maximum
accuracy of 76.4%, 74.6% and 72.5% respectively.
For the batch size of 16, the maximum accuracy rate of 71.0% is achieved without applying
preprocessing technique. After applying the preprocessing techniques, the maximum accuracy is
improved to 76.1%.
12. International Journal of Artificial Intelligence and Applications (IJAIA), Vol.12, No.2, March 2021
12
For the batch size of 16 HE, LTISN and SQI perform very equally well and yield the maximum
accuracy of 76.1%, 72.0% and 71.5% respectively.
For the batch size of 32, the maximum accuracy rate of 68.7% is achieved without applying
preprocessing technique. After applying the preprocessing techniques, the maximum accuracy is
improved to 71.0%.
For the batch size of 32, SQI, GIC, LTISN and DoG perform equally well and yield the
maximum accuracy of 71%, 70.9%, 70.8% and 69.9% respectively.
SQI and HE perform equally well and provide best results for all the batch sizes in both Extended
Yale B and FERET database.
Fig.10.Accuracy using Extended YALE B Database before applying of preprocessing techniques
Fig.11.Accuracy using Extended YALE B Database after application of preprocessing techniques
13. International Journal of Artificial Intelligence and Applications (IJAIA), Vol.12, No.2, March 2021
13
Fig.12.Accuracy using FERET Database before the application of preprocessing techniques.
Fig.13.Accuracy using FERET Database after application of preprocessing techniques.
Table 1 Accuracy rates of Extended Yale B database for various batch sizes before and after application
of preprocessing techniques.
Database Accuracy Rates
Batch Size
4 8 16 32
Extended Yale B (WOAPP) 97.2 97.0 96.8 96.2
DoG 99.1 99.0 98.7 97.9
SQI 99.8 99.4 98.8 98.3
LTISN 99.2 99.3 98.4 98.1
HE 99.7 99.1 99.1 98.3
GIC 99.6 98.9 98.8 98.7
14. International Journal of Artificial Intelligence and Applications (IJAIA), Vol.12, No.2, March 2021
14
Table 2 Accuracy rates of FERET database for various batch sizes before and after application
of preprocessing techniques
Database AccuracyRates
Batch Size
4 8 16 32
FERET(WOAPP) 71.4 71.2 71 68.7
DoG 76.6 72.6 71.3 69.9
SQI 74.9 74.6 71.5 71.0
LTISN 75.6 72.5 72.0 70.8
HE 76.3 76.4 76.1 69.2
GIC 74.2 71.3 71.0 70,9
Data augmentation with the batch size of 128 and number of epochs 250 yields best results,
especially with the preprocessing technique Histogram Equalization. In FERET Database, the
maximum accuracy of 79.6% is achieved.
For the same batch size and number of epochs, LTISN provides almost the same accuracy rate of
78.9%.
DoG, SQI and GIC performs almost equally and provides the accuracy enhancement of 78.2%,
78.3% and 78.1%
Data augmentation Methods does not have much effect in Extended Yale B database. With the
batch size of 128 and number of epochs 250 yields best results, especially with the preprocessing
technique of Histogram Equalization. In Extended Yale B Database, the maximum accuracy of
99.86% is achieved.
For the same batch size and number of epochs, SQI and LTISN provides almost the same
accuracy rate of 99.91% and 99.31%
DoG, HE and GIC performs almost equally and provides the accuracy enhancement of 99.21%,
99.78% and 99.71%
5.1. Number of trainable parameters calculation in Model
The number of trainable parameters is computed for each layer as below. The deep architecture is
as shown in (Fig.14).
C1 Layer
Input image 128×128×1 and having one channel. Number of weights per filter=3×3×1
There are totally eight filters in C1 layer and there is one bias parameter for each filter.
Total number of trainable parameters in C1 Layer is = (3×3×1+1) × 8= 80.
C2 Layer
Input feature maps to C2 layer having 8 channels. Number of weights per filter is 3×3×8.
15. International Journal of Artificial Intelligence and Applications (IJAIA), Vol.12, No.2, March 2021
15
There are 8 such filters in the layer. Hence total number of weights = ((3×3×8) +1) ×8 and there
is one bias for each filter.
Total weights in C2 = ((3×3×8) +1) ×12=584.
S1 Layer
S1 layer is using max pooling operation. Hence no trainable parameter available for this layer.
C3 Layer
Input to C3 has eight feature channels. Number of weights per filter is 3×3×8.
There are 16 such filters in the layer. Hence total number of weights = ((3×3) ×8+1) ×16 and
there is one bias for each filter.
Total weights in C3 = (3×3×8+1) ×16=1168.
C4 Layer
Input to C4 has sixteen feature channels. Number of weights per filter is 3×3×16.
There are 16 such filters in the layer. Hence total number of weights = ((3×3) ×16+1) ×16 and
there is one bias for each filter.
Total weights in C4 = (3×3×16+1) ×16=2320.
S2 Layer
S2 layer is using max pooling operation. Hence no trainable parameter available for this layer.
C5 Layer
Input to C5 has sixteen feature channels. Number of weights per filter is 3×3×16.
There are 32 such filters in the layer. Hence total number of weights = ((3×3) ×16+1) ×32 and
there is one bias for each filter.
Total weights in C4 = (3×3×16+1) ×32=4640.
Fig.14. Deep CNN Model Architecture (BN-Batch Normalization)
16. International Journal of Artificial Intelligence and Applications (IJAIA), Vol.12, No.2, March 2021
16
C6 Layer
Input to C6 has 32 feature channels. Number of weights per filter is 3×3×32
There are 32 such filters in the layer. Hence total number of weights = (3×3 ×32+1) ×32 and
there is one bias for each filter.
Total weights in C4 = (3×3×32+1) ×32=9248.
F1 Layer
In the fully connected layer, the input is having total feature space of size 32×32 and 32. Input
samples to F1 layer=32×32×32
There are 128 neurons in F1 layer and each have a bias.
Total Number of trainable parameters for F1 layer is = (32×32×2+1) ×128=4194432.
Fig.15.Feature map obtained for the batch size 32 and epoch 2
Fig.16.Feature map obtained for the batch size 32 and epoch 5
Fig.17.Feature map obtained for the batch size 32 and epoch 25
17. International Journal of Artificial Intelligence and Applications (IJAIA), Vol.12, No.2, March 2021
17
F2 Layer
The input to F2 layer is the output from F1 layer. It is connected to 28 neurons (since there are 28
classes). So, total number of trainable parameters = (128 +1) ×28= 16512.
Feature maps extracted for different batch size and epochs are as shown in the (Fig.15,16,17).
6. COMPARISON WITH OTHER APPROACHES
In this section, the proposed approach is compared with the various approaches.
Hybrid system, which combines convolutional neural network (CNN) and a Logistic regression
classifier (LRC) yields the accuracy of 80%. A CNN is trained to recognize facial images and
LRC is used to classify the features learned by the convolutional network [26]. The neural
network with the nonlinear theories, such as wavelet theory and fuzzy set is a new research idea,
when applied in face recognition, gives accuracy around 93% in Wavelet-BP [25]. CNN
architecture along with Max-Feature-Map (MFM) could extract compact vital information yields
the accuracy rate of 98% [28]. The popular LeNet-5 architecture in FERET database yields
accuracy rate is 81.25% [27]. The pre-trained CNN model along with the VGG- Face provides
face verification rate accuracy of 86.85% and 83.32% on FRGC and LFW databases respectively
[29].
Table 3. Comparison of Accuracy rates of different approaches.
Sno Architectureused Dataset Accuracy
1 Proposedmethod(C
NN)
Extended Yale
BFERET
99.8%
76.3%
2 VGG+CNN[29] FRGC,LFW 86.85%,
83.32%
3 CNN withMAF
activationfunction
[28]
LFW 98%
4 Wavelet-BP[25] AT&T 93.0%
5 CNN-LRC
[26]
Yale 80%
6 CNN-LENET[27] FERET 81.25%
The proposed method compared with existing approaches are given in Table 3. The proposed
method performs better in comparison to all other approaches in Extended Yale B. In FERET
database, CNN-LENET method performs better than the proposed method.
7. CONCLUSION
In this paper, deep convolution neural network (CNN) is applied for extracting features and
classification. The performance of the network is assessed before and after applying
preprocessing techniques for various batch sizes and epochs on YALE B and FERET dataset.
Adam optimizers with a learning rate of 0.001 is applied in this experiment. The batch sizes used
in this experiment are [4, 8, 16,32];
The optimum batch sizes are chosen using trial and error method. Initially, multiple batch sizes of
[4,8,16,32,64,128] are selected for this experiment. Batch size of 4 with learning rate of0.001 and
18. International Journal of Artificial Intelligence and Applications (IJAIA), Vol.12, No.2, March 2021
18
number of epochs 50 give the best results. Batch sizes of [4,8,16,32] provide consistent and best
results. The batch sizes are normally selected in power of 2. Choosing smaller batch size with
low value of learning rate yields best result.
The results showed the maximum accuracy rate of 97.2% of achieved without using
preprocessing techniques. When the preprocessing techniques are applied, the improved accuracy
rates are achieved up to 99.8%.
The same experiments are conducted in FERET. dataset also. For the frontal face, the maximum
71.4% is achieved without optimization. After applying the preprocessing techniques, the
accuracy rate increased to 76.3%.
The proposed approach performs very well as compared to existing methods.
REFERENCES
[1] S. Setiowati, Zulfanahri, E. L. Franita and I. Ardiyanto, "A review of optimization method in face
recognition: Comparison deep learning and non-deep learning methods” 9th International Conference
on Information Technology and Electrical Engineering (ICITEE), Phuket, pp. 1-6, 2017.
[2] Aleix M. Marttinez and Avinash C. Kak, “PCA versus LDA”. IEEE Transactions on Pattern Analysis
and Machine Intelligence, Vol. 23, No.2, pp. 228-233, 2001.
[3] Rahul Haridas, Jyothi R L “Convolutional Neural Networks: A Comprehensive Survey” International
Journal of Applied Engineering Research Volume 14, Number 3, pp. 780-789, (2019).
[4] Goodfellow, I. J., Bengio, Y., and Courville, A. Deep Learning MIT Press, (2016).
[5] Jason Wang, Luis Perez The Effectiveness of Data Augmentation in Image Classification using Deep
Learning. In: Stanford University research report, 2017.
[6] Di Huang, Caifeng Shan, Mohsen Ardebilian, Yunhong Wang, and Liming Chen “Local Binary
Patterns and Its Application to Facial Image Analysis: A Survey” IEEE Transactions on Systems,
Man and Cybernetics, part C, volume 41, pp 765-781, 2011.
[7] Shailaja A Patil and Dr. P. J. Deore “Face Recognition: A Survey” Informatics Engineering, an
International Journal (IEIJ), Vol.1, No.1, December 2013.
[8] Connor Shorten and Taghi M. Khoshgoftaar “A survey on Image Data Augmentation for Deep
Learning” Jounal of Big Data, - Springer, 2019.
[9] Felix Ellenberger, Claus Lenz “A Non-Technical Survey on Deep Convolutional
Neural Network Architectures” Computer Vision and Pattern Recognition, 2018.
[10] S.V.G. Reddy, K. Thammi Reddy, V. ValliKumari “Optimization of Deep Learning using various
Optimizers, Loss functions and Dropout” International Journal of Recent Technology and
Engineering (IJRTE) Volume-7 Issue-4S2, pp 448-455, December 2018.
[11] SaezTrigueros, D.; Meng, L.; Hartnett, M. Face Recognition: From Traditional to Deep Learning
Methods. J. Theory. Appl. Inf. Technol. 2018, 97, 3332–3342.
[12] Athanasios Voulodimos, Nikolaos Doulamis, AnastasiosDoulamis, and EftychiosProtopapadakis
“Deep Learning for Computer Vision: A Brief Review” Computational Intelligence and
Neuroscience, pp 1-13, 2018.
[13] MdZahangirAlom, Tarek M. Taha, Chris Yakopcic, Stefan Westberg, PahedingSidike,
MstShamimaNasrin, Mahmudul Hasan, Brian C. van Essen, Abdul A. S. Awwal and Vijayan K.
Asari “A State-of-the-Art Survey on Deep Learning Theory and Architectures”. Electronics 2019, 8,
292
[14] WaseemRawat and Zenghui Wang “Deep Convolutional Neural Networks for Image Classification:
A Comprehensive Review” Volume 29, pp 2352-2449 September 2017.
[15] A. F. Agarap. Deep learning using rectified linear units (relu). CoRR, abs/1803.08375, URL
http://arxiv.org/abs/1803.08375 , 2018.
[16] Khan, Asifullah&Sohail, Anabia&Zahoora, Umme& Saeed, Aqsa “A Survey of the Recent
Architectures of Deep Convolutional Neural Networks”, published in Artificial Intelligence Review
(2019).
19. International Journal of Artificial Intelligence and Applications (IJAIA), Vol.12, No.2, March 2021
19
[17] D. H. Hubel and T. N. Wiesel. “Receptive fields of single neurons in the cat’s striate cortex” The
Journal of Physiology, 148(3):574–91, 1959.
[18] Yamashita, R., Nishio, M., Do, R.K.G. et al. “Convolutional neural networks: an overview and
application in radiology” Insights Imaging 9, pp 611–629, 2018.
[19] MamounAlazab, MingJian Tang “Deep Learning Applications for Cyber Security”, Springer, ISBN
978-3- 030-13057-2, May 2019.
[20] N. Aloysius and M. Geetha, "A review on deep convolutional neural networks," 2017 International
Conference on Communication and Signal Processing (ICCSP), Chennai, pp. 0588-0592, doi:
10.1109/ICCSP.2017.8286426, 2017.
[21] Nusrat, Ismoilov and Sung-Bong Jang. “A Comparison of Regularization Techniques in Deep Neural
Networks.” Symmetry 10 (2018): 648.
[22] B. S. Manjunath, R. Chellappa and C. von der Malsburg, "A feature-based approach to face
recognition," Proceedings 1992 IEEE Computer Society Conference on Computer Vision and Pattern
Recognition, Champaign, IL, USA, pp. 373-378, 1992.
[23] Srivastava, Nitish et al. “Dropout: a simple way to prevent neural networks from overfitting.” J.
Mach. Learn. Res. 15, pp 1929-1958, (2014).
[24] Zhao, Qi et al. “Multiactivation Pooling Method in Convolutional Neural Networks for Image
Recognition.” Wireless Communications and Mobile Computing (2018): n. page. Wireless
Communications and Mobile Computing. Web, 2018.
[25] J. Kittler, P. Koppen, et. al, "Conformal Mapping of a 3D Face Representation onto a 2D Image for
CNN Based Face Recognition," 2018 International Conference on Biometrics (ICB), Gold Coast,
QLD, pp. 124-131, 2018.
[26] H. Khalajzadeh, M. Manthouri and M. Teshnehlab, ’Face recognition using convolutional neural
networks and simple logistic classifier’, Advances in intelligent systems and computing, 223, pp 197-
207, (2014).
[27] A.R. syafeeza, M.K. Hani, S.S. Uiw and R. Bakhteri,’ Convolutional neural network for face
recognition with pose and illumination variation’, International journal of engineering and
Technology, 6(1), pp 44-57, (2014).
[28] Y. Xu, F. Liang, G. Zhang and H. Xu, ‘Image intelligent detection based on the Gabor wavelet and
the neural network’, Symmetry, 8(130), (2016).
[29] Wu, Xiang et al. “A Lightened CNN for Deep Face Representation.” ArXiv abs/1511.02683
(2015):n. pag.
[30] Z. Lu, X. Jiang and A. Kot, "Enhance deep learning performance in face recognition,” 2nd
International Conference on Image, Vision and Computing (ICIVC), Chengdu, pp. 244-248, 2017.
[31] Pouyanfar, S., Sadiq, S., Yan, Y., Tian, H., Tao, Y., Reyes, M. P, Iyengar, S. S. A survey on deep
learning: Algorithms, techniques, and applications. ACM Computing Surveys, 51(5), (2018).
[32] Y. Admi, Y. Moses and S. Ulman “Face Recognition: The problem of compensating for changes in
illumination direction”, IEEE Transaction. Pattern Anal. Mach. Intell., vol. 19, pp 721–732, Jul.
1997.
[33] SoodehNikan and M. Ahmadi. “Human Face Recognition under Occlusion using LBP and Entropy
Weighted Voting” 21st International Conference on Pattern Recognition ICPR pp. 1699-1702, 2012.
[34] S. Anila, K. Arulsukanya, K. Divya, C. Kanimozhi, and D. Kavitha “An efficient Preprocessing
Technique for Face Recognition under difficult Lighting Conditions,” in Proc. National Conference
on Emerging Trends in Computer Communication and Informatics (ETCCI-2011), March 10-11,
2011.
[35] Hu Han, Shiguang Shan, Xilin Chen, Wen Gao “A comparative study on illumination preprocessing
in face recognition” Pattern Recognition., Volume 46, Issue 6, Pages 1691-1699, June 2013.
[36] Calvo G., Baruque B., Corchado E. (2013) Study of the Pre-processing Impact in a Facial
Recognition System. In: Pan JS., Polycarpou M.M., Woźniak M., de Carvalho A.C.P.L.F., Quintián
H., Corchado E. (eds) Hybrid Artificial Intelligent Systems. HAIS, Lecture Notes in Computer
Science, vol 8073. Springer, Berlin, Heidelberg, 2013.
[37] Koringa P.A., Mitra S.K., Asari, V.K. Handling Illumination Variation: A Challenge for Face
Recognition. In: Raman B., Kumar S., Roy P., Sen D. (eds) Proceedings of International Conference
on Computer Vision and Image Processing. Advances in Intelligent Systems and Computing, vol 460.
Springer, Singapore, 2017.
20. International Journal of Artificial Intelligence and Applications (IJAIA), Vol.12, No.2, March 2021
20
[38] Dr. A. Sri Krishna, G. Srinivasa Rao and M. Sravya “Contrast Enhancement Techniques using
Histogram Equalization Methods on Color Images with Poor Lightning”, International Journal of
Computer Science, Engineering and Applications (IJCSEA) Vol.3, No.4, August 2013.
[39] P. Suganya, S. Gayathri, N. Mohanapriya “Survey on Image Enhancement Techniques” International
Journal of Computer Applications Technology and Research, Volume 2– Issue 5, 623 - 627, 2013.
[40] Biglari M, Mirzaei F, Ebrahimpour-Komeh H. Illumination invariant face recognition using SQI and
weighted LBP histogram[C]. Pattern Recognition and Image Analysis (PRIA), 2013 First Iranian
Conference on. IEEE, pp 1-7, 2013.
[41] E Krieger, VK Asari, and S Arigela. Color image enhancement of low-resolution images captured in
extreme lighting conditions. In SPIE Sensing Technology and Applications, pages 91200Q– 91200Q.
International Society for Optics and Photonics, 2014.
[42] Anila, S., and N. Devarajan. "Preprocessing technique for face recognition applications under varying
illumination conditions." Global Journal of Computer Science and Technology Volume 12, issue 11,
Version 1, 2012.
[43] NungsanginlaLongkumer, Mukesh Kumar and RohiniSaxena “Contrast Enhancement Techniques
using Histogram Equalization: A Survey” International Journal of Current Engineering and
Technology, Vol.4, No.3, E-ISSN 2277 – 4106, P-ISSN 2347 – 5161 (June 2014).
[44] Ali M. Reza. "Realization of the Contrast Limited Adaptive Histogram Equalization (CLAHE) for
Real-Time Image Enhancement." Journal of VLSI Signal Processing, vol. 38, pp 35-44, 2004.
[45] Porcu, S.; Floris, A.; Atzori, L. Evaluation of Data Augmentation Techniques for Facial Expression
Recognition Systems. Electronics 2020, 9, 1892. https://doi.org/10.3390/electronics9111892.
[46] Zagoruyko, K. Wide residual networks. In BMVC, 2016.
[47] DeVries, T. and Taylor, G. W. Improved regularization of convolutional neural networks with cutout.
arXiv preprint arXiv:1708.04552, 2017.
[48] Y. Bengio, A. Bergeron, N. Boulanger-Lewandowski, T. Breuel, Y. Chherawala, et al. Deep learners
benefit more from out-of-distribution examples. In Proceedings of the Fourteenth International
Conference on Artificial Intelligence and Statistics, pages 164–172, 2011.
[49] R. Wu, S. Yan, Y. Shan, Q. Dang, and G. Sun. Deep image: Scaling up image recognition. arXiv
preprint arXiv:1501.02876, 7(8), 2015.