Most of the currently known methods treat person re-identification task as classification problem and used commonly neural networks. However, these methods used only high-level convolutional feature or to express the feature representation of pedestrians. Moreover, the current data sets for person reidentification is relatively small. Under the limitation of the number of training set, deep convolutional networks are difficult to train adequately. Therefore, it is very worthwhile to introduce auxiliary data sets to help training. In order to solve this problem, this paper propose a novel method of deep transfer learning, and combines the comparison model with the classification model and multi-level fusion of the convolution features on the basis of transfer learning. In a multi-layers convolutional network, the characteristics of each layer of network are the dimensionality reduction of the previous layer of results, but the information of multi-level features is not only inclusive, but also has certain complementarity. We can using the information gap of different layers of convolutional neural networks to extract a better feature expression. Finally, the algorithm proposed in this paper is fully tested on four data sets (VIPeR, CUHK01, GRID and PRID450S). The obtained re-identification results prove the effectiveness of the algorithm.
MULTI-LEVEL FEATURE FUSION BASED TRANSFER LEARNING FOR PERSON RE-IDENTIFICATIONgerogepatton
Most of the currently known methods treat person re-identification task as classification problem and used commonly neural networks. However, these methods used only high-level convolutional feature or to express the feature representation of pedestrians. Moreover, the current data sets for person reidentification is relatively small. Under the limitation of the number of training set, deep convolutional networks are difficult to train adequately. Therefore, it is very worthwhile to introduce auxiliary data sets help training. In order to solve this problem, this paper propose a novel method of deep transfer learning, and combines the comparison model with the classification model and multi-level fusion of the
convolution features on the basis of transfer learning. In a multi-layers convolutional network, the characteristics of each layer of network are the dimensionality reduction of the previous layer of results, but the information of multi-level features is not only inclusive, but also has certain complementarity. We can using the information gap of different layers of convolutional neural networks to extract a better feature expression. Finally, the algorithm proposed in this paper is fully tested on four data sets (VIPeR, CUHK01, GRID and PRID450S). The obtained re-identification results prove the effectiveness of the algorithm.
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.
Automatic gender and age classification has become quite relevant in the rise of social media platforms. However, the existing methods have not been completely successful in achieving this. Through this project, an attempt has been made to determine the gender and age based on a frame of the person. This is done by using deep learning, OpenCV which is capable of processing the real-time frames. This frame is given as input and the predicted gender and age are given as output. It is difficult to predict the exact age of a person using one frame due the facial expressions, lighting, makeup and so on so for this purpose various age ranges are taken, and the predicted age falls in one of them. The Adience dataset is used as it is a benchmark for face photos and includes various real-world imaging conditions like noise, lighting etc.
RunPool: A Dynamic Pooling Layer for Convolution Neural NetworkPutra Wanda
Deep learning (DL) has achieved a significant performance in computer vision problems, mainly in automatic feature extraction and representation. However, it is not easy to determine the best pooling method in a different case study. For instance, experts can implement the best types of pooling in image processing cases, which might not be optimal for various tasks. Thus, it is
required to keep in line with the philosophy of DL. In dynamic neural network architecture, it is not practically possible to find
a proper pooling technique for the layers. It is the primary reason why various pooling cannot be applied in the dynamic and multidimensional dataset. To deal with the limitations, it needs to construct an optimal pooling method as a better option than max pooling and average pooling. Therefore, we introduce a dynamic pooling layer called RunPool to train the convolutional
neuralnetwork(CNN)architecture.RunPoolpoolingisproposedtoregularizetheneuralnetworkthatreplacesthedeterministic
pooling functions. In the final section, we test the proposed pooling layer to address classification problems with online social network (OSN) dataset
MULTI-LEVEL FEATURE FUSION BASED TRANSFER LEARNING FOR PERSON RE-IDENTIFICATIONgerogepatton
Most of the currently known methods treat person re-identification task as classification problem and used commonly neural networks. However, these methods used only high-level convolutional feature or to express the feature representation of pedestrians. Moreover, the current data sets for person reidentification is relatively small. Under the limitation of the number of training set, deep convolutional networks are difficult to train adequately. Therefore, it is very worthwhile to introduce auxiliary data sets help training. In order to solve this problem, this paper propose a novel method of deep transfer learning, and combines the comparison model with the classification model and multi-level fusion of the
convolution features on the basis of transfer learning. In a multi-layers convolutional network, the characteristics of each layer of network are the dimensionality reduction of the previous layer of results, but the information of multi-level features is not only inclusive, but also has certain complementarity. We can using the information gap of different layers of convolutional neural networks to extract a better feature expression. Finally, the algorithm proposed in this paper is fully tested on four data sets (VIPeR, CUHK01, GRID and PRID450S). The obtained re-identification results prove the effectiveness of the algorithm.
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.
Automatic gender and age classification has become quite relevant in the rise of social media platforms. However, the existing methods have not been completely successful in achieving this. Through this project, an attempt has been made to determine the gender and age based on a frame of the person. This is done by using deep learning, OpenCV which is capable of processing the real-time frames. This frame is given as input and the predicted gender and age are given as output. It is difficult to predict the exact age of a person using one frame due the facial expressions, lighting, makeup and so on so for this purpose various age ranges are taken, and the predicted age falls in one of them. The Adience dataset is used as it is a benchmark for face photos and includes various real-world imaging conditions like noise, lighting etc.
RunPool: A Dynamic Pooling Layer for Convolution Neural NetworkPutra Wanda
Deep learning (DL) has achieved a significant performance in computer vision problems, mainly in automatic feature extraction and representation. However, it is not easy to determine the best pooling method in a different case study. For instance, experts can implement the best types of pooling in image processing cases, which might not be optimal for various tasks. Thus, it is
required to keep in line with the philosophy of DL. In dynamic neural network architecture, it is not practically possible to find
a proper pooling technique for the layers. It is the primary reason why various pooling cannot be applied in the dynamic and multidimensional dataset. To deal with the limitations, it needs to construct an optimal pooling method as a better option than max pooling and average pooling. Therefore, we introduce a dynamic pooling layer called RunPool to train the convolutional
neuralnetwork(CNN)architecture.RunPoolpoolingisproposedtoregularizetheneuralnetworkthatreplacesthedeterministic
pooling functions. In the final section, we test the proposed pooling layer to address classification problems with online social network (OSN) dataset
The main objective of this work is the uniting and streamlining of an automatic face detection application and recognition system for video indexing applications. Human identification means the classification of gender which can increase the identification accuracy. So, accurate gender classification algorithms may increase the accuracy of the applications and can reduce its complexity. But, in some applications, some challenges are there such as rotation, gray scale variations that may reduce the accuracy of the application. The main goal of building this module is to understand the values in image, pattern, and array processing with OpenCV for effective processing faces for building pipe-lining, SVM models.
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.
Facial image retrieval on semantic features using adaptive mean genetic algor...TELKOMNIKA JOURNAL
The emergence of larger databases has made image retrieval techniques an essential component and has led to the development of more efficient image retrieval systems. Retrieval can either be content or text-based. In this paper, the focus is on the content-based image retrieval from the FGNET database. Input query images are subjected to several processing techniques in the database before computing the squared Euclidean distance (SED) between them. The images with the shortest Euclidean distance are considered as a match and are retrieved. The processing techniques involve the application of the median modified Weiner filter (MMWF), extraction of the low-level features using histogram-oriented gradients (HOG), discrete wavelet transform (DWT), GIST, and Local tetra pattern (LTrP). Finally, the features are selected using Adaptive Mean Genetic Algorithm (AMGA). In this study, the average PSNR value obtained after applying the Wiener filter was 45.29. The performance of the AMGA was evaluated based on its precision, F-measure, and recall, and the obtained average values were respectively 0.75, 0.692, and 0.66. The performance matrix of the AMGA was compared to those of particle swarm optimization algorithm (PSO) and genetic algorithm (GA) and found to perform better; thus, proving its efficiency.
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
Analysis of Multi-focus Image Fusion Method Based on Laplacian PyramidRajyalakshmi Reddy
This paper presented a simple and efficient algorithm
for multi-focus image fusion, which used a
multiresolution signal decomposition scheme called
Laplacian pyramid method. The principle of Laplacian
pyramid transform is introduced, and based on it the
fusion strategy is described in detail. By analyzing the
experimental results, it showed that this method has
good performance, and the quality of the fused image is
better than the results of other methods
Optimized Neural Network for Classification of Multispectral ImagesIDES Editor
The proposed work involves the multiobjective PSO
based optimization of artificial neural network structure for
the classification of multispectral satellite images. The neural
network is used to classify each image pixel in various land
cove types like vegetations, waterways, man-made structures
and road network. It is per pixel supervised classification using
spectral bands (original feature space). Use of neural network
for classification requires selection of most discriminative
spectral bands and determination of optimal number of nodes
in hidden layer. We propose new methodology based on
multiobjective particle swarm optimization (MOPSO) to
determine discriminative spectral bands and the number of
hidden layer node simultaneously. The result obtained using
such optimized neural network is compared with that of
traditional classifiers like MLC and Euclidean classifier. The
performance of all classifiers is evaluated quantitatively using
Xie-Beni and â indexes. The result shows the superiority of
the proposed method.
ieee projects download, base paper for ieee projects, ieee projects list, ieee projects titles, ieee projects for cse, ieee projects on networking,ieee projects 2012, ieee projects 2013, final year project, computer science final year projects, final year projects for information technology, ieee final year projects, final year students projects, students projects in java, students projects download, students projects in java with source code, students projects architecture, free ieee papers
COMPRESSION BASED FACE RECOGNITION USING DWT AND SVMsipij
The biometric is used to identify a person effectively and employ in almost all applications of day to day
activities. In this paper, we propose compression based face recognition using Discrete Wavelet Transform
(DWT) and Support Vector Machine (SVM). The novel concept of converting many images of single person
into one image using averaging technique is introduced to reduce execution time and memory. The DWT is
applied on averaged face image to obtain approximation (LL) and detailed bands. The LL band coefficients
are given as input to SVM to obtain Support vectors (SV’s). The LL coefficients of DWT and SV’s are fused
based on arithmetic addition to extract final features. The Euclidean Distance (ED) is used to compare test
image features with database image features to compute performance parameters. It is observed that, the
proposed algorithm is better in terms of performance compared to existing algorithms.
Shallow vs. Deep Image Representations: A Comparative Study with Enhancements...CSCJournals
The traditional approach for solving the object recognition problem requires image representations to be first extracted and then fed to a learning model such as an SVM. These representations are handcrafted and heavily engineered by running the object image through a sequence of pipeline steps which requires a good prior knowledge of the problem domain in order to engineer these representations. Moreover, since the classification is done in a separate step, the resultant handcrafted representations are not tuned by the learning model which prevents it from learning complex representations that might would give it more discriminative power. However, in end-to-end deep learning models, image representations along with the classification decision boundary are all learnt directly from the raw data requiring no prior knowledge of the problem domain. These models deeply learn the object image representation hierarchically in multiple layers corresponding to multiple levels of abstraction resulting in representations that are more discriminative and give better results on challenging benchmarks. In contrast to the traditional handcrafted representations, the performance of deep representations improves with the introduction of more data, and more learning layers (more depth) and they perform well on large-scale machine learning problems. The purpose of this study is six fold: (1) review the literature of the pipeline processes used in the previous state-of-the-art codebook model approach for tackling the problem of generic object recognition, (2) Introduce several enhancements in the local feature extraction and normalization steps of the recognition pipeline, (3) compare the enhancements proposed to different encoding methods and contrast them to previous results, (4) experiment with current state-of-the-art deep model architectures used for object recognition, (5) compare between deep representations extracted from the deep learning model and shallow representations handcrafted through the recognition pipeline, and finally, (6) improve the results further by combining multiple different deep learning models into an ensemble and taking the maximum posterior probability.
Comparison Between Levenberg-Marquardt And Scaled Conjugate Gradient Training...CSCJournals
The Internet paved way for information sharing all over the world decades ago and its popularity for distribution of data has spread like a wildfire ever since. Data in the form of images, sounds, animations and videos is gaining users’ preference in comparison to plain text all across the globe. Despite unprecedented progress in the fields of data storage, computing speed and data transmission speed, the demands of available data and its size (due to the increase in both, quality and quantity) continue to overpower the supply of resources. One of the reasons for this may be how the uncompressed data is compressed in order to send it across the network. This paper compares the two most widely used training algorithms for multilayer perceptron (MLP) image compression – the Levenberg-Marquardt algorithm and the Scaled Conjugate Gradient algorithm. We test the performance of the two training algorithms by compressing the standard test image (Lena or Lenna) in terms of accuracy and speed. Based on our results, we conclude that both algorithms were comparable in terms of speed and accuracy. However, the Levenberg- Marquardt algorithm has shown slightly better performance in terms of accuracy (as found in the average training accuracy and mean squared error), whereas the Scaled Conjugate Gradient algorithm faired better in terms of speed (as found in the average training iteration) on a simple MLP structure (2 hidden layers).
Image fusion is a sub field of image processing in which more than one images are fused to create an image where all the objects are in focus. The process of image fusion is performed for multi-sensor and multi-focus images of the same scene. Multi-sensor images of the same scene are captured by different sensors whereas multi-focus images are captured by the same sensor. In multi-focus images, the objects in the scene which are closer to the camera are in focus and the farther objects get blurred. Contrary to it, when the farther objects are focused then closer objects get blurred in the image. To achieve an image where all the objects are in focus, the process of images fusion is performed either in spatial domain or in transformed domain. In recent times, the applications of image processing have grown immensely. Usually due to limited depth of field of optical lenses especially with greater focal length, it becomes impossible to obtain an image where all the objects are in focus. Thus, it plays an important role to perform other tasks of image processing such as image segmentation, edge detection, stereo matching and image enhancement. Hence, a novel feature-level multi-focus image fusion technique has been proposed which fuses multi-focus images. Thus, the results of extensive experimentation performed to highlight the efficiency and utility of the proposed technique is presented. The proposed work further explores comparison between fuzzy based image fusion and neuro fuzzy fusion technique along with quality evaluation indices.
ADOPTING AND IMPLEMENTATION OF SELF ORGANIZING FEATURE MAP FOR IMAGE FUSIONijistjournal
A different image fusion algorithm based on self organizing feature map is proposed in this paper, aiming to produce quality images. Image Fusion is to integrate complementary and redundant information from multiple images of the same scene to create a single composite image that contains all the important features of the original images. The resulting fused image will thus be more suitable for human and machine perception or for further image processing tasks. The existing fusion techniques based on either direct operation on pixels or segments fail to produce fused images of the required quality and are mostly application based. The existing segmentation algorithms become complicated and time consuming when multiple images are to be fused. A new method of segmenting and fusion of gray scale images adopting Self organizing Feature Maps(SOM) is proposed in this paper. The Self Organizing Feature Maps is adopted to produce multiple slices of the source and reference images based on various combination of gray scale and can dynamically fused depending on the application. The proposed technique is adopted and analyzed for fusion of multiple images. The technique is robust in the sense that there will be no loss in information due to the property of Self Organizing Feature Maps; noise removal in the source images done during processing stage and fusion of multiple images is dynamically done to get the desired results. Experimental results demonstrate that, for the quality multifocus image fusion, the proposed method performs better than some popular image fusion methods in both subjective and objective qualities.
Multi-Level Feature Fusion Based Transfer Learning for Person Re-Identificationgerogepatton
Most of the currently known methods treat person re-identification task as classification problem and used
commonly neural networks. However, these methods used only high-level convolutional feature or to
express the feature representation of pedestrians. Moreover, the current data sets for person reidentification is relatively small. Under the limitation of the number of training set, deep convolutional
networks are difficult to train adequately. Therefore, it is very worthwhile to introduce auxiliary data sets
to help training. In order to solve this problem, this paper propose a novel method of deep transfer
learning, and combines the comparison model with the classification model and multi-level fusion of the
convolution features on the basis of transfer learning. In a multi-layers convolutional network, the
characteristics of each layer of network are the dimensionality reduction of the previous layer of results,
but the information of multi-level features is not only inclusive, but also has certain complementarity. We
can using the information gap of different layers of convolutional neural networks to extract a better
feature expression. Finally, the algorithm proposed in this paper is fully tested on four data sets (VIPeR,
CUHK01, GRID and PRID450S). The obtained re-identification results prove the effectiveness of the
algorithm.
The main objective of this work is the uniting and streamlining of an automatic face detection application and recognition system for video indexing applications. Human identification means the classification of gender which can increase the identification accuracy. So, accurate gender classification algorithms may increase the accuracy of the applications and can reduce its complexity. But, in some applications, some challenges are there such as rotation, gray scale variations that may reduce the accuracy of the application. The main goal of building this module is to understand the values in image, pattern, and array processing with OpenCV for effective processing faces for building pipe-lining, SVM models.
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.
Facial image retrieval on semantic features using adaptive mean genetic algor...TELKOMNIKA JOURNAL
The emergence of larger databases has made image retrieval techniques an essential component and has led to the development of more efficient image retrieval systems. Retrieval can either be content or text-based. In this paper, the focus is on the content-based image retrieval from the FGNET database. Input query images are subjected to several processing techniques in the database before computing the squared Euclidean distance (SED) between them. The images with the shortest Euclidean distance are considered as a match and are retrieved. The processing techniques involve the application of the median modified Weiner filter (MMWF), extraction of the low-level features using histogram-oriented gradients (HOG), discrete wavelet transform (DWT), GIST, and Local tetra pattern (LTrP). Finally, the features are selected using Adaptive Mean Genetic Algorithm (AMGA). In this study, the average PSNR value obtained after applying the Wiener filter was 45.29. The performance of the AMGA was evaluated based on its precision, F-measure, and recall, and the obtained average values were respectively 0.75, 0.692, and 0.66. The performance matrix of the AMGA was compared to those of particle swarm optimization algorithm (PSO) and genetic algorithm (GA) and found to perform better; thus, proving its efficiency.
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
Analysis of Multi-focus Image Fusion Method Based on Laplacian PyramidRajyalakshmi Reddy
This paper presented a simple and efficient algorithm
for multi-focus image fusion, which used a
multiresolution signal decomposition scheme called
Laplacian pyramid method. The principle of Laplacian
pyramid transform is introduced, and based on it the
fusion strategy is described in detail. By analyzing the
experimental results, it showed that this method has
good performance, and the quality of the fused image is
better than the results of other methods
Optimized Neural Network for Classification of Multispectral ImagesIDES Editor
The proposed work involves the multiobjective PSO
based optimization of artificial neural network structure for
the classification of multispectral satellite images. The neural
network is used to classify each image pixel in various land
cove types like vegetations, waterways, man-made structures
and road network. It is per pixel supervised classification using
spectral bands (original feature space). Use of neural network
for classification requires selection of most discriminative
spectral bands and determination of optimal number of nodes
in hidden layer. We propose new methodology based on
multiobjective particle swarm optimization (MOPSO) to
determine discriminative spectral bands and the number of
hidden layer node simultaneously. The result obtained using
such optimized neural network is compared with that of
traditional classifiers like MLC and Euclidean classifier. The
performance of all classifiers is evaluated quantitatively using
Xie-Beni and â indexes. The result shows the superiority of
the proposed method.
ieee projects download, base paper for ieee projects, ieee projects list, ieee projects titles, ieee projects for cse, ieee projects on networking,ieee projects 2012, ieee projects 2013, final year project, computer science final year projects, final year projects for information technology, ieee final year projects, final year students projects, students projects in java, students projects download, students projects in java with source code, students projects architecture, free ieee papers
COMPRESSION BASED FACE RECOGNITION USING DWT AND SVMsipij
The biometric is used to identify a person effectively and employ in almost all applications of day to day
activities. In this paper, we propose compression based face recognition using Discrete Wavelet Transform
(DWT) and Support Vector Machine (SVM). The novel concept of converting many images of single person
into one image using averaging technique is introduced to reduce execution time and memory. The DWT is
applied on averaged face image to obtain approximation (LL) and detailed bands. The LL band coefficients
are given as input to SVM to obtain Support vectors (SV’s). The LL coefficients of DWT and SV’s are fused
based on arithmetic addition to extract final features. The Euclidean Distance (ED) is used to compare test
image features with database image features to compute performance parameters. It is observed that, the
proposed algorithm is better in terms of performance compared to existing algorithms.
Shallow vs. Deep Image Representations: A Comparative Study with Enhancements...CSCJournals
The traditional approach for solving the object recognition problem requires image representations to be first extracted and then fed to a learning model such as an SVM. These representations are handcrafted and heavily engineered by running the object image through a sequence of pipeline steps which requires a good prior knowledge of the problem domain in order to engineer these representations. Moreover, since the classification is done in a separate step, the resultant handcrafted representations are not tuned by the learning model which prevents it from learning complex representations that might would give it more discriminative power. However, in end-to-end deep learning models, image representations along with the classification decision boundary are all learnt directly from the raw data requiring no prior knowledge of the problem domain. These models deeply learn the object image representation hierarchically in multiple layers corresponding to multiple levels of abstraction resulting in representations that are more discriminative and give better results on challenging benchmarks. In contrast to the traditional handcrafted representations, the performance of deep representations improves with the introduction of more data, and more learning layers (more depth) and they perform well on large-scale machine learning problems. The purpose of this study is six fold: (1) review the literature of the pipeline processes used in the previous state-of-the-art codebook model approach for tackling the problem of generic object recognition, (2) Introduce several enhancements in the local feature extraction and normalization steps of the recognition pipeline, (3) compare the enhancements proposed to different encoding methods and contrast them to previous results, (4) experiment with current state-of-the-art deep model architectures used for object recognition, (5) compare between deep representations extracted from the deep learning model and shallow representations handcrafted through the recognition pipeline, and finally, (6) improve the results further by combining multiple different deep learning models into an ensemble and taking the maximum posterior probability.
Comparison Between Levenberg-Marquardt And Scaled Conjugate Gradient Training...CSCJournals
The Internet paved way for information sharing all over the world decades ago and its popularity for distribution of data has spread like a wildfire ever since. Data in the form of images, sounds, animations and videos is gaining users’ preference in comparison to plain text all across the globe. Despite unprecedented progress in the fields of data storage, computing speed and data transmission speed, the demands of available data and its size (due to the increase in both, quality and quantity) continue to overpower the supply of resources. One of the reasons for this may be how the uncompressed data is compressed in order to send it across the network. This paper compares the two most widely used training algorithms for multilayer perceptron (MLP) image compression – the Levenberg-Marquardt algorithm and the Scaled Conjugate Gradient algorithm. We test the performance of the two training algorithms by compressing the standard test image (Lena or Lenna) in terms of accuracy and speed. Based on our results, we conclude that both algorithms were comparable in terms of speed and accuracy. However, the Levenberg- Marquardt algorithm has shown slightly better performance in terms of accuracy (as found in the average training accuracy and mean squared error), whereas the Scaled Conjugate Gradient algorithm faired better in terms of speed (as found in the average training iteration) on a simple MLP structure (2 hidden layers).
Image fusion is a sub field of image processing in which more than one images are fused to create an image where all the objects are in focus. The process of image fusion is performed for multi-sensor and multi-focus images of the same scene. Multi-sensor images of the same scene are captured by different sensors whereas multi-focus images are captured by the same sensor. In multi-focus images, the objects in the scene which are closer to the camera are in focus and the farther objects get blurred. Contrary to it, when the farther objects are focused then closer objects get blurred in the image. To achieve an image where all the objects are in focus, the process of images fusion is performed either in spatial domain or in transformed domain. In recent times, the applications of image processing have grown immensely. Usually due to limited depth of field of optical lenses especially with greater focal length, it becomes impossible to obtain an image where all the objects are in focus. Thus, it plays an important role to perform other tasks of image processing such as image segmentation, edge detection, stereo matching and image enhancement. Hence, a novel feature-level multi-focus image fusion technique has been proposed which fuses multi-focus images. Thus, the results of extensive experimentation performed to highlight the efficiency and utility of the proposed technique is presented. The proposed work further explores comparison between fuzzy based image fusion and neuro fuzzy fusion technique along with quality evaluation indices.
ADOPTING AND IMPLEMENTATION OF SELF ORGANIZING FEATURE MAP FOR IMAGE FUSIONijistjournal
A different image fusion algorithm based on self organizing feature map is proposed in this paper, aiming to produce quality images. Image Fusion is to integrate complementary and redundant information from multiple images of the same scene to create a single composite image that contains all the important features of the original images. The resulting fused image will thus be more suitable for human and machine perception or for further image processing tasks. The existing fusion techniques based on either direct operation on pixels or segments fail to produce fused images of the required quality and are mostly application based. The existing segmentation algorithms become complicated and time consuming when multiple images are to be fused. A new method of segmenting and fusion of gray scale images adopting Self organizing Feature Maps(SOM) is proposed in this paper. The Self Organizing Feature Maps is adopted to produce multiple slices of the source and reference images based on various combination of gray scale and can dynamically fused depending on the application. The proposed technique is adopted and analyzed for fusion of multiple images. The technique is robust in the sense that there will be no loss in information due to the property of Self Organizing Feature Maps; noise removal in the source images done during processing stage and fusion of multiple images is dynamically done to get the desired results. Experimental results demonstrate that, for the quality multifocus image fusion, the proposed method performs better than some popular image fusion methods in both subjective and objective qualities.
Multi-Level Feature Fusion Based Transfer Learning for Person Re-Identificationgerogepatton
Most of the currently known methods treat person re-identification task as classification problem and used
commonly neural networks. However, these methods used only high-level convolutional feature or to
express the feature representation of pedestrians. Moreover, the current data sets for person reidentification is relatively small. Under the limitation of the number of training set, deep convolutional
networks are difficult to train adequately. Therefore, it is very worthwhile to introduce auxiliary data sets
to help training. In order to solve this problem, this paper propose a novel method of deep transfer
learning, and combines the comparison model with the classification model and multi-level fusion of the
convolution features on the basis of transfer learning. In a multi-layers convolutional network, the
characteristics of each layer of network are the dimensionality reduction of the previous layer of results,
but the information of multi-level features is not only inclusive, but also has certain complementarity. We
can using the information gap of different layers of convolutional neural networks to extract a better
feature expression. Finally, the algorithm proposed in this paper is fully tested on four data sets (VIPeR,
CUHK01, GRID and PRID450S). The obtained re-identification results prove the effectiveness of the
algorithm.
Comparison of convolutional neural network models for user’s facial recognitionIJECEIAES
This paper compares well-known convolutional neural networks (CNN) models for facial recognition. For this, it uses its database created from two registered users and an additional category of unknown persons. Eight different base models of convolutional architectures were compared by transfer of learning, and two additional proposed models called shallow CNN and shallow directed acyclic graph with CNN (DAG-CNN), which are architectures with little depth (six convolution layers). Within the tests with the database, the best results were obtained by the GoogLeNet and ResNet-101 models, managing to classify 100% of the images, even without confusing people outside the two users. However, in an additional real-time test, in which one of the users had his style changed, the models that showed the greatest robustness in this situation were the Inception and the ResNet-101, being able to maintain constant recognition. This demonstrated that the networks of greater depth manage to learn more detailed features of the users' faces, unlike those of shallower ones; their learning of features is more generalized. Declare the full term of an abbreviation/acronym when it is mentioned for the first time.
Image segmentation and classification tasks in computer vision have proven to be highly effective using neural networks, specifically Convolutional Neural Networks (CNNs). These tasks have numerous
practical applications, such as in medical imaging, autonomous driving, and surveillance. CNNs are capable
of learning complex features directly from images and achieving outstanding performance across several
datasets. In this work, we have utilized three different datasets to investigate the efficacy of various preprocessing and classification techniques in accurssedately segmenting and classifying different structures
within the MRI and natural images. We have utilized both sample gradient and Canny Edge Detection
methods for pre-processing, and K-means clustering have been applied to segment the images. Image
augmentation improves the size and diversity of datasets for training the models for image classification
Image segmentation and classification tasks in computer vision have proven to be highly effective using neural networks, specifically Convolutional Neural Networks (CNNs). These tasks have numerous
practical applications, such as in medical imaging, autonomous driving, and surveillance. CNNs are capable
of learning complex features directly from images and achieving outstanding performance across several
datasets. In this work, we have utilized three different datasets to investigate the efficacy of various preprocessing and classification techniques in accurssedately segmenting and classifying different structures
within the MRI and natural images. We have utilized both sample gradient and Canny Edge Detection
methods for pre-processing, and K-means clustering have been applied to segment the images. Image
augmentation improves the size and diversity of datasets for training the models for image classification.
This work highlights transfer learning’s effectiveness in image classification using CNNs and VGG 16 that
provides insights into the selection of pre-trained models and hyper parameters for optimal performance.
We have proposed a comprehensive approach for image segmentation and classification, incorporating preprocessing techniques, the K-means algorithm for segmentation, and employing deep learning models such
as CNN and VGG 16 for classification.
Human Re-identification with Global and Local Siamese Convolution Neural NetworkTELKOMNIKA JOURNAL
Human re-identification is an important task in surveillance system to determine whether the same human re-appears in multiple cameras with disjoint views. Mostly, appearance based approaches are used to perform human re-identification task because they are less constrained than biometric based approaches. Most of the research works apply hand-crafted feature extractors and then simple matching methods are used. However, designing a robust and stable feature requires expert knowledge and takes time to tune the features. In this paper, we propose a global and local structure of Siamese Convolution Neural Network which automatically extracts features from input images to perform human re-identification task. Besides, most of the current human re-identification tasks in single-shot approaches do not consider occlusion issue due to lack of tracking information. Therefore, we apply a decision fusion technique to combine global and local features for occlusion cases in single-shot approaches.
Noisy image enhancements using deep learning techniquesIJECEIAES
This article explores the application of deep learning techniques to improve the accuracy of feature enhancements in noisy images. A multitasking convolutional neural network (CNN) learning model architecture has been proposed that is trained on a large set of annotated images. Various techniques have been used to process noisy images, including the use of data augmentation, the application of filters, and the use of image reconstruction techniques. As a result of the experiments, it was shown that the proposed model using deep learning methods significantly improves the accuracy of object recognition in noisy images. Compared to single-tasking models, the multi-tasking model showed the superiority of this approach in performing multiple tasks simultaneously and saving training time. This study confirms the effectiveness of using multitasking models using deep learning for object recognition in noisy images. The results obtained can be applied in various fields, including computer vision, robotics, automatic driving, and others, where accurate object recognition in noisy images is a critical component.
Global-local attention with triplet loss and label smoothed cross-entropy for...IAESIJAI
Person re-identification (Person Re-ID) is a research direction on tracking and identifying people in surveillance camera systems with non-overlapping camera perspectives. Despite much research on this topic, there are still some practical problems that Person Re-ID has not yet solved, in reality, human objects can easily be obscured by obstructions such as other people, trees, luggage, umbrellas, signs, cars, motorbikes. In this paper, we propose a multi-branch deep learning network architecture. In which one branch is for the representation of global features and two branches are for the representation of local features. Dividing the input image into small parts and changing the number of parts between the two branches helps the model to represent the features better. In addition, we add an attention module to the ResNet50 backbone that enhances important human characteristics and eliminates irrelevant information. To improve robustness, the model is trained by combining triplet loss and label smoothing cross-entropy loss (LSCE). Experiments are carried out on datasets Market1501, and duke multi-target multi-camera (DukeMTMC) datasets, our method achieved 96.04% rank-1, 88,11% mean average precision (mAP) on the Market1501 dataset, and 88.78% rank-1, 78,6% mAP on the DukeMTMC dataset. This method achieves performance better than some state-of-the-art methods.
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.
Noise-robust classification with hypergraph neural networknooriasukmaningtyas
This paper presents a novel version of hypergraph neural network method. This method is utilized to solve the noisy label learning problem. First, we apply the PCA dimensional reduction technique to the feature matrices of the image datasets in order to reduce the “noise” and the redundant features in the feature matrices of the image datasets and to reduce the runtime constructing the hypergraph of the hypergraph neural network method. Then, the classic graph based semisupervised learning method, the classic hypergraph based semi-supervised learning method, the graph neural network, the hypergraph neural network, and our proposed hypergraph neural network are employed to solve the noisy label learning problem. The accuracies of these five methods are evaluated and compared. Experimental results show that the hypergraph neural network methods achieve the best performance when the noise level increases. Moreover, the hypergraph neural network methods are at least as good as the graph neural network.
Machine learning based augmented reality for improved learning application th...IJECEIAES
Detection of objects and their location in an image are important elements of current research in computer vision. In May 2020, Meta released its state-ofthe-art object-detection model based on a transformer architecture called detection transformer (DETR). There are several object-detection models such as region-based convolutional neural network (R-CNN), you only look once (YOLO) and single shot detectors (SSD), but none have used a transformer to accomplish this task. These models mentioned earlier, use all sorts of hyperparameters and layers. However, the advantages of using a transformer pattern make the architecture simple and easy to implement. In this paper, we determine the name of a chemical experiment through two steps: firstly, by building a DETR model, trained on a customized dataset, and then integrate it into an augmented reality mobile application. By detecting the objects used during the realization of an experiment, we can predict the name of the experiment using a multi-class classification approach. The combination of various computer vision techniques with augmented reality is indeed promising and offers a better user experience.
FACE EXPRESSION RECOGNITION USING CONVOLUTION NEURAL NETWORK (CNN) MODELS ijgca
This paper proposes the design of a Facial Expression Recognition (FER) system based on deep
convolutional neural network by using three model. In this work, a simple solution for facial expression
recognition that uses a combination of algorithms for face detection, feature extraction and classification
is discussed. The proposed method uses CNN models with SVM classifier and evaluates them, these models
are Alex-net model, VGG-16 model and Res-Net model. Experiments are carried out on the Extended
Cohn-Kanada (CK+) datasets to determine the recognition accuracy for the proposed FER system. In this
study the accuracy of AlexNet model compared with Vgg16 model and ResNet model. The result show that
AlexNet model achieved the best accuracy (88.2%) compared to other models.
Semi-Supervised Method of Multiple Object Segmentation with a Region Labeling...sipij
Efficient and efficient multiple object segmentation is an important task in computer vision and object recognition. In this work; we address a method to effectively discover a user’s concept when multiple objects of interest are involved in content based image retrieval. The proposed method incorporate a framework for multiple object retrieval using semi-supervised method of similar region merging and flood fill which models the spatial and appearance relations among image pixels. To improve the effectiveness of similarity based region merging we propose a new similarity based object retrieval. The users only need to roughly indicate the after which steps desired objects contour is obtained during the automatic merging of similar regions. A novel similarity based region merging mechanism is proposed to guide the merging process with the help of mean shift technique and objects detection using region labeling and flood fill. A region R is merged with its adjacent regions Q if Q has highest similarity with Q (using Bhattacharyya descriptor) among all Q’s adjacent regions. The proposed method automatically merges the regions that are initially segmented through mean shift technique, and then effectively extracts the object contour by merging all similar regions. Extensive experiments are performed on 12 object classes (224 images total) show promising results.
Real-time eyeglass detection using transfer learning for non-standard facial...IJECEIAES
The aim of this paper is to build a real-time eyeglass detection framework based on deep features present in facial or ocular images, which serve as a prime factor in forensics analysis, authentication systems and many more. Generally, eyeglass detection methods were executed using cleaned and fine-tuned facial datasets; it resulted in a well-developed model, but the slightest deviation could affect the performance of the model giving poor results on real-time non-standard facial images. Therefore, a robust model is introduced which is trained on custom non-standard facial data. An Inception V3 architecture based pre-trained convolutional neural network (CNN) is used and fine-tuned using model hyper-parameters to achieve a high accuracy and good precision on non-standard facial images in real-time. This resulted in an accuracy score of about 99.2% and 99.9% for training and testing datasets respectively in less amount of time thereby showing the robustness of the model in all conditions.
Similar to MULTI-LEVEL FEATURE FUSION BASED TRANSFER LEARNING FOR PERSON RE-IDENTIFICATION (20)
Saudi Arabia stands as a titan in the global energy landscape, renowned for its abundant oil and gas resources. It's the largest exporter of petroleum and holds some of the world's most significant reserves. Let's delve into the top 10 oil and gas projects shaping Saudi Arabia's energy future in 2024.
CFD Simulation of By-pass Flow in a HRSG module by R&R Consult.pptxR&R Consult
CFD analysis is incredibly effective at solving mysteries and improving the performance of complex systems!
Here's a great example: At a large natural gas-fired power plant, where they use waste heat to generate steam and energy, they were puzzled that their boiler wasn't producing as much steam as expected.
R&R and Tetra Engineering Group Inc. were asked to solve the issue with reduced steam production.
An inspection had shown that a significant amount of hot flue gas was bypassing the boiler tubes, where the heat was supposed to be transferred.
R&R Consult conducted a CFD analysis, which revealed that 6.3% of the flue gas was bypassing the boiler tubes without transferring heat. The analysis also showed that the flue gas was instead being directed along the sides of the boiler and between the modules that were supposed to capture the heat. This was the cause of the reduced performance.
Based on our results, Tetra Engineering installed covering plates to reduce the bypass flow. This improved the boiler's performance and increased electricity production.
It is always satisfying when we can help solve complex challenges like this. Do your systems also need a check-up or optimization? Give us a call!
Work done in cooperation with James Malloy and David Moelling from Tetra Engineering.
More examples of our work https://www.r-r-consult.dk/en/cases-en/
Overview of the fundamental roles in Hydropower generation and the components involved in wider Electrical Engineering.
This paper presents the design and construction of hydroelectric dams from the hydrologist’s survey of the valley before construction, all aspects and involved disciplines, fluid dynamics, structural engineering, generation and mains frequency regulation to the very transmission of power through the network in the United Kingdom.
Author: Robbie Edward Sayers
Collaborators and co editors: Charlie Sims and Connor Healey.
(C) 2024 Robbie E. Sayers
Democratizing Fuzzing at Scale by Abhishek Aryaabh.arya
Presented at NUS: Fuzzing and Software Security Summer School 2024
This keynote talks about the democratization of fuzzing at scale, highlighting the collaboration between open source communities, academia, and industry to advance the field of fuzzing. It delves into the history of fuzzing, the development of scalable fuzzing platforms, and the empowerment of community-driven research. The talk will further discuss recent advancements leveraging AI/ML and offer insights into the future evolution of the fuzzing landscape.
Quality defects in TMT Bars, Possible causes and Potential Solutions.PrashantGoswami42
Maintaining high-quality standards in the production of TMT bars is crucial for ensuring structural integrity in construction. Addressing common defects through careful monitoring, standardized processes, and advanced technology can significantly improve the quality of TMT bars. Continuous training and adherence to quality control measures will also play a pivotal role in minimizing these defects.
NO1 Uk best vashikaran specialist in delhi vashikaran baba near me online vas...Amil Baba Dawood bangali
Contact with Dawood Bhai Just call on +92322-6382012 and we'll help you. We'll solve all your problems within 12 to 24 hours and with 101% guarantee and with astrology systematic. If you want to take any personal or professional advice then also you can call us on +92322-6382012 , ONLINE LOVE PROBLEM & Other all types of Daily Life Problem's.Then CALL or WHATSAPP us on +92322-6382012 and Get all these problems solutions here by Amil Baba DAWOOD BANGALI
#vashikaranspecialist #astrologer #palmistry #amliyaat #taweez #manpasandshadi #horoscope #spiritual #lovelife #lovespell #marriagespell#aamilbabainpakistan #amilbabainkarachi #powerfullblackmagicspell #kalajadumantarspecialist #realamilbaba #AmilbabainPakistan #astrologerincanada #astrologerindubai #lovespellsmaster #kalajaduspecialist #lovespellsthatwork #aamilbabainlahore#blackmagicformarriage #aamilbaba #kalajadu #kalailam #taweez #wazifaexpert #jadumantar #vashikaranspecialist #astrologer #palmistry #amliyaat #taweez #manpasandshadi #horoscope #spiritual #lovelife #lovespell #marriagespell#aamilbabainpakistan #amilbabainkarachi #powerfullblackmagicspell #kalajadumantarspecialist #realamilbaba #AmilbabainPakistan #astrologerincanada #astrologerindubai #lovespellsmaster #kalajaduspecialist #lovespellsthatwork #aamilbabainlahore #blackmagicforlove #blackmagicformarriage #aamilbaba #kalajadu #kalailam #taweez #wazifaexpert #jadumantar #vashikaranspecialist #astrologer #palmistry #amliyaat #taweez #manpasandshadi #horoscope #spiritual #lovelife #lovespell #marriagespell#aamilbabainpakistan #amilbabainkarachi #powerfullblackmagicspell #kalajadumantarspecialist #realamilbaba #AmilbabainPakistan #astrologerincanada #astrologerindubai #lovespellsmaster #kalajaduspecialist #lovespellsthatwork #aamilbabainlahore #Amilbabainuk #amilbabainspain #amilbabaindubai #Amilbabainnorway #amilbabainkrachi #amilbabainlahore #amilbabaingujranwalan #amilbabainislamabad
TECHNICAL TRAINING MANUAL GENERAL FAMILIARIZATION COURSEDuvanRamosGarzon1
AIRCRAFT GENERAL
The Single Aisle is the most advanced family aircraft in service today, with fly-by-wire flight controls.
The A318, A319, A320 and A321 are twin-engine subsonic medium range aircraft.
The family offers a choice of engines
Courier management system project report.pdfKamal Acharya
It is now-a-days very important for the people to send or receive articles like imported furniture, electronic items, gifts, business goods and the like. People depend vastly on different transport systems which mostly use the manual way of receiving and delivering the articles. There is no way to track the articles till they are received and there is no way to let the customer know what happened in transit, once he booked some articles. In such a situation, we need a system which completely computerizes the cargo activities including time to time tracking of the articles sent. This need is fulfilled by Courier Management System software which is online software for the cargo management people that enables them to receive the goods from a source and send them to a required destination and track their status from time to time.
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.
About
Indigenized remote control interface card suitable for MAFI system CCR equipment. Compatible for IDM8000 CCR. Backplane mounted serial and TCP/Ethernet communication module for CCR remote access. IDM 8000 CCR remote control on serial and TCP protocol.
• Remote control: Parallel or serial interface.
• Compatible with MAFI CCR system.
• Compatible with IDM8000 CCR.
• Compatible with Backplane mount serial communication.
• Compatible with commercial and Defence aviation CCR system.
• Remote control system for accessing CCR and allied system over serial or TCP.
• Indigenized local Support/presence in India.
• Easy in configuration using DIP switches.
Technical Specifications
Indigenized remote control interface card suitable for MAFI system CCR equipment. Compatible for IDM8000 CCR. Backplane mounted serial and TCP/Ethernet communication module for CCR remote access. IDM 8000 CCR remote control on serial and TCP protocol.
Key Features
Indigenized remote control interface card suitable for MAFI system CCR equipment. Compatible for IDM8000 CCR. Backplane mounted serial and TCP/Ethernet communication module for CCR remote access. IDM 8000 CCR remote control on serial and TCP protocol.
• Remote control: Parallel or serial interface
• Compatible with MAFI CCR system
• Copatiable with IDM8000 CCR
• Compatible with Backplane mount serial communication.
• Compatible with commercial and Defence aviation CCR system.
• Remote control system for accessing CCR and allied system over serial or TCP.
• Indigenized local Support/presence in India.
Application
• Remote control: Parallel or serial interface.
• Compatible with MAFI CCR system.
• Compatible with IDM8000 CCR.
• Compatible with Backplane mount serial communication.
• Compatible with commercial and Defence aviation CCR system.
• Remote control system for accessing CCR and allied system over serial or TCP.
• Indigenized local Support/presence in India.
• Easy in configuration using DIP switches.
MULTI-LEVEL FEATURE FUSION BASED TRANSFER LEARNING FOR PERSON RE-IDENTIFICATION
1. International Journal of Artificial Intelligence and Applications (IJAIA), Vol.10, No.3, May 2019
DOI: 10.5121/ijaia.2019.10302 13
MULTI-LEVEL FEATURE FUSION BASED TRANSFER
LEARNING FOR PERSON RE-IDENTIFICATION
Yingzhi Chen and Tianqi Yang
Department of Computer Science, Jinan University, China
ABSTRACT
Most of the currently known methods treat person re-identification task as classification problem and used
commonly neural networks. However, these methods used only high-level convolutional feature or to
express the feature representation of pedestrians. Moreover, the current data sets for person re-
identification is relatively small. Under the limitation of the number of training set, deep convolutional
networks are difficult to train adequately. Therefore, it is very worthwhile to introduce auxiliary data sets
to help training. In order to solve this problem, this paper propose a novel method of deep transfer
learning, and combines the comparison model with the classification model and multi-level fusion of the
convolution features on the basis of transfer learning. In a multi-layers convolutional network, the
characteristics of each layer of network are the dimensionality reduction of the previous layer of results,
but the information of multi-level features is not only inclusive, but also has certain complementarity. We
can using the information gap of different layers of convolutional neural networks to extract a better
feature expression. Finally, the algorithm proposed in this paper is fully tested on four data sets (VIPeR,
CUHK01, GRID and PRID450S). The obtained re-identification results prove the effectiveness of the
algorithm.
KEYWORDS
Person re-identification; transfer learning; multi-level feature fusion & deep learning
1. INTRODUCTION
Person re-identification is an important area of research in computer vision. The goal of person
re-identification is to identify the same person from the pictures of candidates, and the probe and
gallery are capture different cameras. It is worth noting that the shooting areas of these cameras
often do not intersect. In recent years, person re-identification systems have become more widely
used in criminal investigation and law enforcement agencies. The complete system consists of
person detection, person tracking and person retrieval. Each part is complex, and the three parts
of the computer vision field have evolved independent tasks. The person re-identification task
content studied in this paper is roughly the same as the person retrieval task. It consists of two
steps: feature extraction and distance matching between candidate sets. In the first step, there are
generally two difficulties of person re-identification: 1) Detail features such as face, fingerprint
and iris are not stable due to the resolution and angle of view of person image. 2) Person pictures
are collected by cameras with different perspectives. When person are captured at different
positions, the body posture and angle will be different. Person remain unchanged under different
cameras in terms of overall characteristics such as clothing, shape and skin. Thus, more
distinguishing features are nesseary. For the distance matching step, given a target picture and a
candidate picture set, the correct picture in the candidate set should be closest to the target picture
in the feature space. The distance measure of the picture can be supervised or unsupervised. This
article focuses on supervised learning. The basic process of pedestrian recognition is shown in
Figure 1.
2. International Journal of Artificial Intelligence and Applications (IJAIA), Vol.10, No.3, May 2019
14
Figure 1. the process of person re-identification
Since 2012, Krizhevsky et al. [1] used the convolutional network to win the championship in the
ILSVRC (ImageNet Large Scale Visual Recognition Competition)-2012 image classification
competition. Deep learning has become one of the research hotspots in many field [24, 25, 26].
Deep learning method has a good effect. At present, person re-identification task also begins to
adopt the deep learning method, and has achieved good experimental results. Li et al. [2] first
proposed a FPNN (Filter pairing neural network) model based on convolutional network, which
first uses convolution and pooling operation turns the picture into a series of local features, and
then the image is divided into multiple regions for matching, which used a unified neural network
framework to handle the effects of background noise, lighting, and occlusion. Ahmed et al. [3]
proposed an enhanced neural network framework that takes advantage of the proximity
differences in input picture pairs and picture blocks to simplify features. Compared with manual
feature extraction, the process of deep learning to extract features is more direct and convenient,
and it is no longer necessary to manually design features based on person images. The
characteristics of deep learning can be automatically adapted to the influence of angle,
illumination and occlusion in person re-identification. This is because deep learning has a large
number of learnable parameters. By continuously adjusting parameters, a target feature related to
a specific target task can be obtained. Despite the advantages of deep learning, it relies on the
number of samples, especially the number of samples in the classification task, and the small-
scale data is difficult to train a suitable neural network. Transfer learning is different from general
machine learning, is aims to learn knowledge form one domain and use the knowledge on other
different domain. Transfer learning had wide applications in computer vision [21, 22, 23]. The
training data and test data of machine learning are generally from the same task, and the training
data of the transfer learning and the source of the test data are inconsistent, there is a large gap
between the training data and the target data. How to reduce this distribution difference is the key
concern of transfer learning. When applying transfer learning in a neural network, there are
usually two ways to reduce the distribution difference. One is to select the auxiliary data as close
as possible to the target data, and the other is to perform secondary training on the target data, that
is, fine-tuning. The transfer learning process in deep learning has two steps: pre-training and fine-
tuning. First, the task of learning in the past is called the source task, the data is the source data,
the task to prepare for learning is called the target task, and the data is the target data. Pre-training
3. International Journal of Artificial Intelligence and Applications (IJAIA), Vol.10, No.3, May 2019
15
means that the model is trained on the source data in advance to complete the source task. Fine-
tuning refers to the second training of the pre-trained model on the target data to complete the
target task. From the perspective of initialization, pre-training is equivalent to having the model
parameters have an initial value that is more in line with the target task, and the fine-tuning
process is training on this initialization. In the case where the target data and source data are more
consistently distributed, there are two main advantages of fine-tuning: (1) Save time costs and use
pre-trained models without reinitializing the network for training. (2)With better generalization
ability, the pre-trained model is generally trained on large data sets. Combined with the data set of
the target task, it is equivalent to using additional auxiliary data to complete the target task.
At present, the mainstream method of using a multi-layer convolution network is to convolve the
feature map layer by layer from the input picture, and then input the final layer convolution result
as a final feature of the pedestrian picture to the classifier. Most of gestures, clothing and
accessories of person are very similar, using the most prominent features of the unity does not
make the most of the full use of person information. This article considers this issue, so in this
chapter, we explore the multi-level features of convolutional networks. Convolutional networks
use only three simple operations: convolution, pooling, and full connection. Convolution and
pooling alternate to form the main part of the network. These operations perform a large number
of complex operations, extracting abstract features, making the interior of the product network is
more like a "black box." In recent years, many researchers have used convolution visualization to
explain the mechanism of convolutional networks. A lot of work has been done on the features
learned by convolutional networks. Convolution visualization can be traced back to AlexNet [1]
proposed by Krizhevshy et al., which directly visualizes the convolution kernel of the first
convolutional layer of the network, and displays the value of the convolution kernel as a picture.
The response strength of the product is visually visible, but the smaller convolution kernel of this
method, such as 3×3 and 5×1, is difficult to apply because the resolution of the small convolution
kernel after conversion to a picture is very low, and it is difficult to obtain meaningful results.
The work of Grishick et al. [4] then marked the area of the AlexNet that responded strongly to the
input data. Zeiler et al. [5] used the same convolution kernel transposition as the network itself as
a new convolution operation, taking the convolved feature map as input, de-convolve the same
form of image as the original input, and improving the visualization based on the visualization
results of AlexNet, and achieved good results in related tasks. We using multi-level feature fusion
of convolutional neural network and twice transfer learning to learn more information of person.
2. OUR METHOD
2.1. Two-Step Transfer Learning
Figure 2 shows the two-step transfer learning process. It is divided into three phases. The specific
settings are shown in Table 1. After the auxiliary data set is determined and the training model
obtains a basic model, the basic model needs to be fine-tuned on the target data to obtain the final
model. The whole process is divided into two phases: the pre-training phase and the fine-tuning
phase. In the pre-training phase, our neural network model is consistent with the models used in
other general visual classifications. The data set is not specially processed. The network structure
used the ResNet [6] structure and the loss function is the classification loss. The architecture of
ResNet is shown in Table 1.
4. International Journal of Artificial Intelligence and Applications (IJAIA), Vol.10, No.3, May 2019
16
Table 1. The architecture of ResNet
Layer Kernel Channel Repeat Pooling
CNN 7 × 7 64 1 3x3, max
CNN
1 × 1
3 × 3
1 × 1
64
64
256
3 none
CNN
1 × 1
3 × 3
1 × 1
128
128
512
4 none
CNN
1 × 1
3 × 3
1 × 1
256
256
1024
6 none
CNN
1 × 1
3 × 3
1 × 1
512
512
2048
3 none
Fully-connect none none 1 average
In the fine-tuning stage, for person re-identification, it can be divided into two models to fine-
tune: classification model and verification model. Because the test data set is too small during the
one-step transfer process and spans three types of data sets, the simple classification model and
the verification model can't achieve good results, so this paper proposed a new joint model
combining the two models. Thus the general process of two-step deep transfer learning is divided
into pre-training, fine-tuning and feature extraction. Since the tasks for pre-training are different
from the tasks for fine-tuning and feature extraction, the network parameters and structure of
deep learning will retain most of them and change a small part, so that pre-training learning can
retain knowledge, while fine-tuning the network can learn new knowledge, the changed part is
often the last layer of the full connection, because the last layer of the fully connected layer of
neuron network in the classification task and the number of species and the last layer of
information strong correlation with task. In this paper, the classification model is used in the pre-
training phase. In the fine-tuning phase, the same number of neurons as the pedestrian re-
identification identity is used to form the full-connection layer instead of the original layer. The
activation function is softmax. After the desired category is output, the intersection is calculated.
And the output of the last layer of convolution is used as the input of the above new full
connection, and the second is the characteristic input of the contrast loss, thereby calculating the
cross entropy while calculating the contrast loss, and finally adding the two losses as the final
loss, using this loss to fine-tune the pre-trained network. In the last phase, we extract the feature
maps of convolutional layers. The hyper-parameters of pre-training same as [6]. And in fine-
tuning, the hyper-parameters are shown in Table 2.
5. International Journal of Artificial Intelligence and Applications (IJAIA), Vol.10, No.3, May 2019
17
Figure 2. Two-step Transfer Learning
Table 2. The Hyper-Parameters in Fine-Tuning
Learning rate
Decay of
weight
Decay of
learning rate
Optimization Batch
0.0003 0.0005 0.1 Adam[27] 32
2.1. Multi-level Feature Fusion
The convolutional network has the characteristics of parameter sharing and local sensing. In
addition, it can be found that the characteristics of the multi-layer convolution network have the
following special properties:
(1) On the image, the convolution is calculated by weighting and summing a pixel and its
adjacent pixels. The intuitive interpretation of this method is to change the value of this pixel
according to the information around the pixel. If the value of the convolution kernel is 0 except
for the target pixel being 1, then the information of each pixel output is the same as the input
information; if the value of the convolution kernel is different, the output pixel will contain
information of other pixels, assuming the convolution kernel value is [1,0,1], then for each pixel
of the input, the output will be the sum of the pixel values on both sides of the pixel, that is, the
information of the pixel is filtered out and replaced by the pixel information on both sides, so the
volume The product feature is the filtering result of the input of this layer. After learning, it will
filter the noise that is meaningless to the task result, which can be regarded as the abstraction of
the input data.
(2) In a network consisting of a full-convolutional, the output characteristics of each layer are the
result of convolution operations on the input of this layer, and the convolution has an abstraction
effect, also assuming that the value of each layer of convolution kernel is [1,0,1], the pixel value
of the first layer convolution result is the sum of the pixel values on both sides, the new pixel
value filters out the original middle pixel, and the second layer convolution result is the filtering
of the first layer result, nth The layer output is the filtered result of the n-1th layer output, and this
abstraction exists in each layer of convolution. As the number of network layers grows deeper,
6. International Journal of Artificial Intelligence and Applications (IJAIA), Vol.10, No.3, May 2019
18
the convolution feature has more and more abstract features, that is, it has progressive abstract
features.
(3) Under ideal conditions, the multi-level convolution network is well trained and the optimal
solution is obtained. Then the last layer is characterized by all the information that affects the
recognition effect. However, the training of neural network is a non-convex optimization
problem. The actual situation often cannot reach the ideal state. Therefore, convolution filtering
the input data noise information will also filter out some meaningful information, which will
affect the recognition effect. Multi-level features will have some complementarity. It can be
summarized as a multi-level feature with three characteristics: abstraction, progressiveness and
complementarity.
Given an input x, the output of the convolutional layer is:
𝑥 𝑛
= 𝑓(𝑊 𝑛
𝑥 𝑛−1
+ 𝑏 𝑛), 𝑛 ∈ (1, 𝑁) (1)
Where f denotes the activation function. After the fine-tuning of the multi-layer convolution
network, the traditional approach is to input x into the network, and then extract the output of the
last layer of convolution as xN
a feature to represent x, we have some complementarity according
to multi-level features. We extract the output of the multi-layer convolution and fuse it into X as
the feature of the person picture in the form of splicing, the form of X is:
X = [𝑥1
, 𝑥2
, … , 𝑥 𝑁
] (2)
That is:
X = [𝑓(𝑊1
𝑥0
+ 𝑏1), 𝑓(𝑊2
𝑥1
+ 𝑏2), … , 𝑓(𝑊 𝑁
𝑥 𝑁−1
+ 𝑏 𝑁)] (3)
The input image is a colour image, that is, there are three colour channels of R, G, and B. In each
layer, the number of channels according to different feature patterns of the channel parameters
changes accordingly, and each feature map size of the random multi-level feature is different, and
each feature map is doubled by a pooling layer with a step size of 2×2. This section uses the
pooled feature map as a feature representation, so that the characteristics of each layer are due to
the dimension. The proportion of each layer in the fusion feature is different. Because the degree
of influence of each layer on the recognition result is unknown, this section extracts each layer
after the feature is extracted, so that the feature dimensions of each layer after pooling are
consistent. Use 𝑝𝑜𝑜𝑙 𝑛
to represent the nth layer of pooling, then:
X = [𝑝𝑜𝑜𝑙1
(𝑓(𝑊1
𝑥0
+ 𝑏1)), 𝑝𝑜𝑜𝑙2
(𝑓(𝑊2
𝑥1
+ 𝑏2)), … , 𝑝𝑜𝑜𝑙 𝑁
(𝑓(𝑊 𝑁
𝑥 𝑁−1
+ 𝑏 𝑁))]
(4)
The multi-level fusion method is shown in Figure 3. After the person image input into the
network, the output of the selected pooling layer is extracted. This output is originally flowed as
an input to the next layer network. This algorithm uses it as a method. Part of the expression of
the overall characteristics of pedestrians, the progressiveness proposed in the above summary is
slowly progressive in multi-layer convolution networks, and the degree of increase in the degree
of abstraction between each adjacent convolutional layer is not very large. That is, the closer the
information is to the convolution feature on the network structure, the more similar the
information is, smaller the information difference is. There are information gap in each
convolutional layers.
7. International Journal of Artificial Intelligence and Applications (IJAIA), Vol.10, No.3, May 2019
19
Figure 3. Feature Fusion
3. EXPERIMENTS
3.1. Datasets
Experiments are conducted in the following three datasets:
VIPeR: The VIPeR dataset [7] was collected from the outdoor environment by the University of
California Computer Vision Lab and contained 632 pedestrians under two cameras. The same
pedestrian consists of two pictures. The size of all the images is 128x48, and the angles, postures
and illumination of the two pictures of the same pedestrian are not the same. Most of the two
pictures have a viewing angle of about 90 degrees, among which pedestrians More will carry a
backpack and the appearance of the backpack changes at different angles.
CUHK01: The CUHK01 data set [8] is collected from two cameras on the campus of the Chinese
University of Hong Kong. It consists of a total of 971 pedestrians, and each pedestrian has two
pictures under the same camera, and each pedestrian is photographed by two cameras, that is,
each pedestrian contains 4 pictures with a total of 3884 images, because it is shot on campus. The
captured pedestrians carry more school bags.
PRID450S: The PRID450S dataset [9] is based on the PRID2011 dataset and contains 450 pairs
of pedestrian images for a total of 900 images. All images are captured by two fixed cameras with
no cross-field of view. The original captured data is in video format. The image data is obtained
by cropping. The resolution of the cropped images is not consistent and the shooting location is
on an outdoor street. Most pedestrians pose. It is the side, and because the shooting time has a
certain interval, the pedestrian will move and cause the pedestrian background to be different.
GRID: The GRID data set [10] is a smaller data set that is collected from eight disjoint cameras at
the underground station. It contains 250 pedestrian image pairs, that is, each pedestrian has two
images captured from different cameras, and also contains 775 additional images, which do not
belong to 250 pedestrians, and therefore exist as interference. GRID has a large change in
8. International Journal of Artificial Intelligence and Applications (IJAIA), Vol.10, No.3, May 2019
20
pedestrian attitude, color and illumination, and the resolution of the picture is low. The
underground parking lot is dense, so the picture contains more than one pedestrian. Other
pedestrians are used as the background, and there are cases where they are blocked by other
pedestrians.
3.2. Experimental results
This summary compares the multi-level fusion feature on the four data and the deep and
handcraft fusion feature with the mainstream method from 15 to 18 years, as shown in Table 3. In
the case of fine-tuning, in terms of Rank-1 scores, the deep and handcraft features achieved the
best results on the VIPeR, CUHK01, and PRID450S datasets, with the best results compared to
other publicly available results increased by 9.67%, 3.9%, 4.7%, respectively.. The Rank-1 score
achieved the second best result on the GRID dataset. The multi-level fusion feature on VIPeR and
CUHK01 also achieved the best score compared with other public results. The GRID dataset
contains a large proportion of unidentified information. Person pictures lead to a more specific
identification. Although this method does not achieve the optimal level on Rank-1, it has
achieved a good results. The comparison methods selected in this paper include deep learning
methods and manual feature methods. Although the former results are often better than the latter,
the latter will still achieve better results in some tests. Under the non-fine tuning setting, the
fusion feature of this paper has achieved the best results on the VIPeR, CUHK01 and PRID450S
datasets compared to the mainstream methods in the past three years. The non-fine tuning result
on CUHK01 is higher than the fine-tuning result of 1.6% on PRID450S. The ratio is 0.1% higher
than the fine-tuning result, and the other two data sets are lower than the fine-tuning result. This
verifies the multi-level feature fusion and the effectiveness of the manual-depth feature fusion,
and the improvement effect of the joint model on migration learning.
Table 3. Overall Performance
VIPeR CUHK01 PRID450S GRID
Method
Rank-
1
Rank
-10
Rank
-20
Rank
-1
Rank
-10
Rank
-20
Ran
k-1
Rank-
10
Rank-
20
Rank
-1
Rank-
10
Rank-
20
LOMO+X
QDA[11]
ICCV
15
40.00
80.5
1
91.0
8
63.2
1
- - 61.4 91.0 95.3
18.9
6
52.56 62.24
KHPCA[1
2]
ICPR
16
39.4 85.1 93.5 - - - 52.2 92.8 94.4 - - -
LSSCDL[1
3]
CVP
R16
42.7 84.3 91.9
65.9
7
- - 60.5 88.6 93.6 22.4 51.3 61.2
GOG+XQ
DA[14]
CVP
R16
49.7 88.7 94.5 67.3 91.8 95.9 68.4 94.5 97.8 24.7 58.4 69.0
DNS[15]
CVP
R16
51.17
90.5
1
95.9
2
69.0
9
91.7
7
95.3
9
- - - - - -
DLPAR[1
6]
ICCV
17
48.7 85.1 93.0 75.0 95.7 97.7 - - - - - -
JLML[17]
IJCAI
17
50.2 84.3 91.6 76.7 95.6 98.1 - - -
37.5
0
69.40 77.40
SSM[18]
CVP
R17
53.73
91.4
9
96.0
8
72.9
8
96.7
6
99.1
1
- - -
27.2
0
61.12 70.56
EDPRMLS
[19]
CVP
R18
50.10
84.3
5
- - - - - - - - - -
GCT[20]
AAA
I18
49.4 87.2 94.0 61.9 87.6 92.8 58.4 84.3 89.8 - - -
Ours(Deep) 59.1 91.1 96.1 77.9 95.6 98.0 59.0 88.8 93.9 20.8 49.1 58.1
Ours(Fusion &
Fine Tune)
63.4 93.3 97.1 80.6 96.6 98.5 73.1 95.0 98.3 27.4 57.6 69.0
9. International Journal of Artificial Intelligence and Applications (IJAIA), Vol.10, No.3, May 2019
21
Ours(Fusion &
No-Fine Tune)
57.6 92.2 96.8 82.2 97.6 99.1 73.2 96.0 99.0 25.4 57.4 69.8
4. CONCLUSIONS
In order to apply transfer learning more effectively, this paper proposes a joint model based on
the comparison model and classification model, and applies the joint model to the fine-tuning
phase of migration learning. Therefore, the model can learn the category-related features as well
as the features that can make the same class features closer in the feature space and the different
class features are farther away. This paper visualizes and discusses the results of different layers
of multi-layer convolutional networks, and summarizes the characteristics of multi-level features
with abstraction, progressiveness and certain complementarity. According to the characteristics of
certain complementarity, a method of merging multi-level features is proposed, which expands
the feature expression of convolutional networks. This paper selects four widely used public
datasets, namely VIPeR, GRID, PRID450S and CUHK01, to verify the algorithm
comprehensively. The experimental results show that the joint model and feature fusion have
obvious performance improvement compared with single feature. By comparing the algorithm
results published by researchers in the field of pedestrian recognition in recent years, it is proved
that the method has good recognition accuracy.
REFERENCES
[1] Krizhevsky, A., Sutskever, I., & Hinton, G. E. (2012). Imagenet classification with deep
convolutional neural networks. In Advances in neural information processing systems (pp. 1097-
1105).
[2] Li, W., Zhao, R., Xiao, T., & Wang, X. (2014). Deepreid: Deep filter pairing neural network for
person re-identification. In Proceedings of the IEEE Conference on Computer Vision and Pattern
Recognition (pp. 152-159).
[3] Ahmed, E., Jones, M., & Marks, T. K. (2015). An improved deep learning architecture for person re-
identification. In Proceedings of the IEEE conference on computer vision and pattern recognition (pp.
3908-3916).
[4] Girshick, R., Donahue, J., Darrell, T., & Malik, J. (2014). Rich feature hierarchies for accurate object
detection and semantic segmentation. In Proceedings of the IEEE conference on computer vision and
pattern recognition (pp. 580-587).
[5] Zeiler, M. D., & Fergus, R. (2014, September). Visualizing and understanding convolutional
networks. In European conference on computer vision (pp. 818-833). springer, Cham.
[6] He, K., Zhang, X., Ren, S., & Sun, J. (2016). Deep residual learning for image recognition. In
Proceedings of the IEEE conference on computer vision and pattern recognition (pp. 770-778).
[7] Gray, D., & Tao, H. (2008, October). Viewpoint invariant pedestrian recognition with an ensemble of
localized features. In European conference on computer vision (pp. 262-275). Springer, Berlin,
Heidelberg.
[8] Li, W., Zhao, R., & Wang, X. (2012, November). Human reidentification with transferred metric
learning. In Asian conference on computer vision (pp. 31-44). Springer, Berlin, Heidelberg.
[9] Roth, P. M., Hirzer, M., Köstinger, M., Beleznai, C., & Bischof, H. (2014). Mahalanobis distance
learning for person re-identification. In Person re-identification (pp. 247-267). Springer, London.
10. International Journal of Artificial Intelligence and Applications (IJAIA), Vol.10, No.3, May 2019
22
[10] Van De Weijer, J., Schmid, C., Verbeek, J., & Larlus, D. (2009). Learning color names for real-world
applications. IEEE Transactions on Image Processing, 18(7), 1512-1523.
[11] Liao, S., Hu, Y., Zhu, X., & Li, S. Z. (2015). Person re-identification by local maximal occurrence
representation and metric learning. In Proceedings of the IEEE conference on computer vision and
pattern recognition (pp. 2197-2206).
[12] Prates, R. F., & Schwartz, W. R. (2016, December). Kernel hierarchical PCA for person re-
identification. In 2016 23rd International Conference on Pattern Recognition (ICPR) (pp. 2091-2096).
IEEE.
[13] Zhang, Y., Li, B., Lu, H., Irie, A., & Ruan, X. (2016). Sample-specific svm learning for person re-
identification. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition
(pp. 1278-1287).
[14] Matsukawa, T., Okabe, T., Suzuki, E., & Sato, Y. (2016). Hierarchical gaussian descriptor for person
re-identification. In Proceedings of the IEEE Conference on Computer Vision and Pattern
Recognition (pp. 1363-1372).
[15] Zhang, L., Xiang, T., & Gong, S. (2016). Learning a discriminative null space for person re-
identification. In Proceedings of the IEEE conference on computer vision and pattern recognition (pp.
1239-1248).
[16] Zhao, L., Li, X., Zhuang, Y., & Wang, J. (2017). Deeply-learned part-aligned representations for
person re-identification. In Proceedings of the IEEE International Conference on Computer Vision
(pp. 3219-3228).
[17] Li, W., Zhu, X., & Gong, S. (2017). Person re-identification by deep joint learning of multi-loss
classification. arXiv preprint arXiv:1705.04724.
[18] Bai, S., Bai, X., & Tian, Q. (2017). Scalable person re-identification on supervised smoothed
manifold. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (pp.
2530-2539).
[19] Guo, Y., & Cheung, N. M. (2018). Efficient and deep person re-identification using multi-level
similarity. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (pp.
2335-2344).
[20] Zhou, Q., Fan, H., Zheng, S., Su, H., Li, X., Wu, S., & Ling, H. (2018, April). Graph correspondence
transfer for person re-identification. In Thirty-Second AAAI Conference on Artificial Intelligence.
[21] Saenko, K., Kulis, B., Fritz, M., & Darrell, T. (2010, September). Adapting visual category models to
new domains. In European conference on computer vision (pp. 213-226). Springer, Berlin,
Heidelberg.
[22] Gopalan, R., Li, R., & Chellappa, R. (2011, November). Domain adaptation for object recognition:
An unsupervised approach. In 2011 international conference on computer vision (pp. 999-1006).
IEEE.
[23] Hoffman, J., Guadarrama, S., Tzeng, E. S., Hu, R., Donahue, J., Girshick, R., ... & Saenko, K. (2014).
LSDA: Large scale detection through adaptation. In Advances in Neural Information Processing
Systems (pp. 3536-3544).
[24] Penha. A & Penha. D. (2018, March). Home Appliance Identification for Nilm Systems Based on
Deep Neural Networks. International Journal of Artificial Intelligence & Applications (IJAIA), (pp.
69-80).
11. International Journal of Artificial Intelligence and Applications (IJAIA), Vol.10, No.3, May 2019
23
[25] Salahat. S & Awad. M. (2017, March). Short-Term Forecasting of Electricity Consumption in
Palestine Using Artificial Neural Networks. International Journal of Artificial Intelligence &
Applications (IJAIA), (pp. 11-21).
[26] Artificial Neural Networks. International Journal of Artificial Intelligence & Applications (IJAIA),
(pp. 17-29).
[27] Kingma, D. P., & Ba, J. (2014). Adam: A method for stochastic optimization. arXiv preprint
arXiv:1412.6980.