Content-based image retrieval (CBIR) uses the content features for
retrieving and searching the images in a given large database. Earlier,
different hand feature descriptor designs are researched based on cues that
are visual such as shape, colour, and texture used to represent these images.
Although, deep learning technologies have widely been applied as an
alternative to designing engineering that is dominant for over a decade. The
features are automatically learnt through the data. This research work
proposes integrated dual deep convolutional neural network (IDD-CNN),
IDD-CNN comprises two distinctive CNN, first CNN exploits the features
and further custom CNN is designed for exploiting the custom features.
Moreover, a novel directed graph is designed that comprises the two blocks
i.e. learning block and memory block which helps in finding the similarity
among images; since this research considers the large dataset, an optimal
strategy is introduced for compact features. Moreover, IDD-CNN is
evaluated considering the two distinctive benchmark datasets the oxford
dataset considering mean average precision (mAP) metrics and comparative
analysis shows IDD-CNN outperforms the other existing model.
Content-based image retrieval based on corel dataset using deep learningIAESIJAI
A popular technique for retrieving images from huge and unlabeled image databases are content-based-image-retrieval (CBIR). However, the traditional information retrieval techniques do not satisfy users in terms of time consumption and accuracy. Additionally, the number of images accessible to users are growing due to web development and transmission networks. As the result, huge digital image creation occurs in many places. Therefore, quick access to these huge image databases and retrieving images like a query image from these huge image collections provides significant challenges and the need for an effective technique. Feature extraction and similarity measurement are important for the performance of a CBIR technique. This work proposes a simple but efficient deep-learning framework based on convolutional-neural networks (CNN) for the feature extraction phase in CBIR. The proposed CNN aims to reduce the semantic gap between low-level and high-level features. The similarity measurements are used to compute the distance between the query and database image features. When retrieving the first 10 pictures, an experiment on the Corel-1K dataset showed that the average precision was 0.88 with Euclidean distance, which was a big step up from the state-of-the-art approaches.
This document discusses content-based image mining techniques for image retrieval. It provides an overview of image mining, describing how image mining goes beyond content-based image retrieval by aiming to discover significant patterns in large image collections according to user queries. The document reviews several existing image mining techniques, including those using color histograms, texture analysis, clustering algorithms like k-means, and association rule mining. It discusses challenges in developing universal image retrieval methods and proposes combining low-level visual features with high-level semantic features. Overall, the document surveys the state of the art in content-based image mining and retrieval.
This document discusses content-based image mining techniques for image retrieval. It provides an overview of image mining, describing how image mining goes beyond content-based image retrieval by aiming to discover significant patterns in large image collections according to user queries. The document reviews several existing image mining techniques, including those using color histograms, texture analysis, clustering algorithms like k-means, and association rule mining. It discusses challenges in developing universal image retrieval methods and proposes combining low-level visual features with high-level semantic features. Overall, the document surveys the state of the art in content-based image mining and retrieval.
A SURVEY ON CONTENT BASED IMAGE RETRIEVAL USING MACHINE LEARNINGIRJET Journal
This document provides a literature review of recent research on content-based image retrieval using machine learning techniques. It summarizes 8 research papers that used approaches like convolutional neural networks, color histograms, deep learning, hashing functions and more to extract image features and retrieve similar images from databases. The goal of content-based image retrieval is to find images that are semantically similar to a query image based on visual features.
Batik image retrieval using convolutional neural networkTELKOMNIKA JOURNAL
This paper presents a simple technique for performing Batik image retrieval using the Convolutional Neural Network (CNN) approach. Two CNN models, i.e. supervised and unsupervised learning approach, are considered to perform end-to-end feature extraction in order to describe the content of Batik image. The distance metrics measure the similarity between the query and target images in database based on the feature generated from CNN architecture. As reported in the experimental section, the proposed supervised CNN model achieves better performance compared to unsupervised CNN in the Batik image retrieval system. In addition, image feature composed from the proposed CNN model yields better performance compared to that of the handcrafted feature descriptor. Yet, it demonstrates the superiority performance of deep learning-based approach in the Batik image retrieval system.
This document provides a comprehensive review of recent developments in content-based image retrieval and feature extraction. It discusses various low-level visual features used for image retrieval, including color, texture, shape, and spatial features. It also reviews approaches that fuse low-level features and use local features. Machine learning and deep learning techniques for content-based image retrieval are also summarized. The document concludes by discussing open challenges and directions for future research in this area.
Content Based Image Retrieval (CBIR) aims at retrieving the images from the database based on the user query which is visual form rather than the traditional text form. The applications of CBIR extend from surveillance to remote sensing, medical imaging to weather forecasting, and security systems to historical research and so on. Though extensive research is made on content based image retrieval in the spatial domain, we have most images in the internet which is JPEG compressed which pushes the need for image retrieval in the compressed domain itself rather than decoding it to raw format before comparison and retrieval. This research addresses the need to retrieve the images from the database based on the features extracted from the compressed domain along with the application of genetic algorithm in improving the retrieval results. The research focuses on various features and their levels of impact on improving the precision and recall parameters of the CBIR system. Our experimentation results also indicate that the CBIR features in compressed domain along with the genetic algorithm usage improves the results considerably when compared with the literature techniques.
A Survey On: Content Based Image Retrieval Systems Using Clustering Technique...IJMIT JOURNAL
This document summarizes various content-based image retrieval techniques using clustering methods for large datasets. It discusses clustering algorithms like K-means, hierarchical clustering, graph-based clustering and a proposed hybrid divide-and-conquer K-means method. The hybrid method uses hierarchical and divide-and-conquer approaches to improve K-means performance for high dimensional datasets. Content-based image retrieval relies on automatically extracted visual features like color, texture and shape for image classification and retrieval.
Content-based image retrieval based on corel dataset using deep learningIAESIJAI
A popular technique for retrieving images from huge and unlabeled image databases are content-based-image-retrieval (CBIR). However, the traditional information retrieval techniques do not satisfy users in terms of time consumption and accuracy. Additionally, the number of images accessible to users are growing due to web development and transmission networks. As the result, huge digital image creation occurs in many places. Therefore, quick access to these huge image databases and retrieving images like a query image from these huge image collections provides significant challenges and the need for an effective technique. Feature extraction and similarity measurement are important for the performance of a CBIR technique. This work proposes a simple but efficient deep-learning framework based on convolutional-neural networks (CNN) for the feature extraction phase in CBIR. The proposed CNN aims to reduce the semantic gap between low-level and high-level features. The similarity measurements are used to compute the distance between the query and database image features. When retrieving the first 10 pictures, an experiment on the Corel-1K dataset showed that the average precision was 0.88 with Euclidean distance, which was a big step up from the state-of-the-art approaches.
This document discusses content-based image mining techniques for image retrieval. It provides an overview of image mining, describing how image mining goes beyond content-based image retrieval by aiming to discover significant patterns in large image collections according to user queries. The document reviews several existing image mining techniques, including those using color histograms, texture analysis, clustering algorithms like k-means, and association rule mining. It discusses challenges in developing universal image retrieval methods and proposes combining low-level visual features with high-level semantic features. Overall, the document surveys the state of the art in content-based image mining and retrieval.
This document discusses content-based image mining techniques for image retrieval. It provides an overview of image mining, describing how image mining goes beyond content-based image retrieval by aiming to discover significant patterns in large image collections according to user queries. The document reviews several existing image mining techniques, including those using color histograms, texture analysis, clustering algorithms like k-means, and association rule mining. It discusses challenges in developing universal image retrieval methods and proposes combining low-level visual features with high-level semantic features. Overall, the document surveys the state of the art in content-based image mining and retrieval.
A SURVEY ON CONTENT BASED IMAGE RETRIEVAL USING MACHINE LEARNINGIRJET Journal
This document provides a literature review of recent research on content-based image retrieval using machine learning techniques. It summarizes 8 research papers that used approaches like convolutional neural networks, color histograms, deep learning, hashing functions and more to extract image features and retrieve similar images from databases. The goal of content-based image retrieval is to find images that are semantically similar to a query image based on visual features.
Batik image retrieval using convolutional neural networkTELKOMNIKA JOURNAL
This paper presents a simple technique for performing Batik image retrieval using the Convolutional Neural Network (CNN) approach. Two CNN models, i.e. supervised and unsupervised learning approach, are considered to perform end-to-end feature extraction in order to describe the content of Batik image. The distance metrics measure the similarity between the query and target images in database based on the feature generated from CNN architecture. As reported in the experimental section, the proposed supervised CNN model achieves better performance compared to unsupervised CNN in the Batik image retrieval system. In addition, image feature composed from the proposed CNN model yields better performance compared to that of the handcrafted feature descriptor. Yet, it demonstrates the superiority performance of deep learning-based approach in the Batik image retrieval system.
This document provides a comprehensive review of recent developments in content-based image retrieval and feature extraction. It discusses various low-level visual features used for image retrieval, including color, texture, shape, and spatial features. It also reviews approaches that fuse low-level features and use local features. Machine learning and deep learning techniques for content-based image retrieval are also summarized. The document concludes by discussing open challenges and directions for future research in this area.
Content Based Image Retrieval (CBIR) aims at retrieving the images from the database based on the user query which is visual form rather than the traditional text form. The applications of CBIR extend from surveillance to remote sensing, medical imaging to weather forecasting, and security systems to historical research and so on. Though extensive research is made on content based image retrieval in the spatial domain, we have most images in the internet which is JPEG compressed which pushes the need for image retrieval in the compressed domain itself rather than decoding it to raw format before comparison and retrieval. This research addresses the need to retrieve the images from the database based on the features extracted from the compressed domain along with the application of genetic algorithm in improving the retrieval results. The research focuses on various features and their levels of impact on improving the precision and recall parameters of the CBIR system. Our experimentation results also indicate that the CBIR features in compressed domain along with the genetic algorithm usage improves the results considerably when compared with the literature techniques.
A Survey On: Content Based Image Retrieval Systems Using Clustering Technique...IJMIT JOURNAL
This document summarizes various content-based image retrieval techniques using clustering methods for large datasets. It discusses clustering algorithms like K-means, hierarchical clustering, graph-based clustering and a proposed hybrid divide-and-conquer K-means method. The hybrid method uses hierarchical and divide-and-conquer approaches to improve K-means performance for high dimensional datasets. Content-based image retrieval relies on automatically extracted visual features like color, texture and shape for image classification and retrieval.
A deep locality-sensitive hashing approach for achieving optimal image retri...IJECEIAES
The proposed method achieves optimal image retrieval through a deep locality-sensitive hashing approach. It extracts both low-level visual features and high-level semantic features from different layers of a CNN model. Locality-sensitive hashing is applied to the features to generate hash codes, which are then used to quickly retrieve similar images based on Hamming distance. Experimental results on CIFAR-10 and NUS-WIDE datasets show the proposed method outperforms other hash-based image retrieval methods in terms of accuracy and retrieval time.
End-to-end deep auto-encoder for segmenting a moving object with limited tra...IJECEIAES
The document proposes two end-to-end deep auto-encoder approaches for segmenting moving objects from surveillance videos when limited training data is available. The first approach uses transfer learning with a pre-trained VGG-16 model as the encoder and its transposed architecture as the decoder. The second approach uses a multi-depth auto-encoder with convolutional and upsampling layers. Both approaches apply data augmentation techniques like PCA and traditional methods to increase the training data size. The models are trained and evaluated on the CDnet2014 dataset, achieving better performance than other models trained with limited data.
The advents in this technological era have resulted into enormous pool of information. This information is
stored at multiple places globally, in multiple formats. This article highlights a methodology for extracting
the video lectures delivered by experts in the domain of Computer Science by using Generalized Gamma
Mixture Model. The feature extraction is based on the DCT transformations. In order to propose the model,
the data set is pooled from the YouTube video lectures in the domain of Computer Science. The outputs
generated are evaluated using Precision and Recall.
Precision face image retrieval by extracting the face features and comparing ...prjpublications
This document describes a proposed method for improving content-based face image retrieval. The method uses two orthogonal techniques: attribute-enhanced sparse coding and attribute-embedded inverted indexing. Attribute-enhanced sparse coding exploits global features to construct semantic codewords offline. Attribute-embedded inverted indexing considers local query image features in a binary signature to efficiently retrieve images. By combining these techniques, the method reduces errors and achieves better face image extraction from databases compared to existing content-based retrieval systems. It works by extracting features from the query image, matching them to database images, and returning ranked results.
Content-based Image Retrieval System for an Image Gallery Search Application IJECEIAES
Content-based image retrieval is a process framework that applies computer vision techniques for searching and managing large image collections more efficiently. With the growth of large digital image collections triggered by rapid advances in electronic storage capacity and computing power, there is a growing need for devices and computer systems to support efficient browsing, searching, and retrieval for image collections. Hence, the aim of this project is to develop a content-based image retrieval system that can be implemented in an image gallery desktop application to allow efficient browsing through three different search modes: retrieval by image query, retrieval by facial recognition, and retrieval by text or tags. In this project, the MPEG-7-like Powered Localized Color and Edge Directivity Descriptor is used to extract the feature vectors of the image database and the facial recognition system is built around the Eigenfaces concept. A graphical user interface with the basic functionality of an image gallery application is also developed to implement the three search modes. Results show that the application is able to retrieve and display images in a collection as thumbnail previews with high retrieval accuracy and medium relevance and the computational requirements for subsequent searches were significantly reduced through the incorporation of text-based image retrieval as one of the search modes. All in all, this study introduces a simple and convenient way of offline image searches on desktop computers and provides a stepping stone to future content-based image retrieval systems built for similar purposes.
RECURRENT FEATURE GROUPING AND CLASSIFICATION MODEL FOR ACTION MODEL PREDICTI...IJDKP
Content based retrieval has an advantage of higher prediction accuracy as compared to tagging based approach. However, the complexity in its representation and classification approach, results in lower processing accuracy and computation overhead. The correlative nature of the feature data are un-explored in the conventional modeling, where all the data features are taken as a set of feature values to give a decision. The recurrent feature class attribute is observed for the feature regrouping in action model prediction. In this paper a co-relative information, bounding grouping approach is suggested for action model prediction in CBMR application. The co-relative recurrent feature mapping results in faster retrieval process as compared to the conventional retrieval system.
RECURRENT FEATURE GROUPING AND CLASSIFICATION MODEL FOR ACTION MODEL PREDICTI...IJDKP
Content based retrieval has an advantage of higher prediction accuracy as compared to tagging based approach. However, the complexity in its representation and classification approach, results in lower processing accuracy and computation overhead. The correlative nature of the feature data are un-explored in the conventional modeling, where all the data features are taken as a set of feature values to give a decision. The recurrent feature class attribute is observed for the feature regrouping in action model prediction. In this paper a co-relative information, bounding grouping approach is suggested for action model prediction in CBMR application. The co-relative recurrent feature mapping results in faster retrieval process as compared to the conventional retrieval system.
Mining of images using retrieval techniqueseSAT Journals
Abstract Today’s world is digital. Use of different social websites is become a part of our day to day life. These social websites have acquired great popularity because of their user-friendly features and their content. Nowadays internet and mobile networks are widely used everywhere and so use of images. Because of evolution in hardware as well as software it is possible to store large amount of multimedia data. In short, now a day it becomes easy to store huge amount of images by using image processing techniques. As number of images and databases are increasing day by day, there is a need for new image retrieval techniques that should be fulfilled. Image mining is derived from data mining which is a method to extract information from digital image. The purpose of this paper is to present a review of the various image mining techniques used in different applications as image retrieval, Matching, Pattern recognition etc.given by different researchers. An information or knowledge can be extracted from the image by using image mining technique. This paper proposes the survey in image retrieval techniques. Image retrieval and data mining have huge applications in the sector of image processing, pattern recognition, image matching, image mining, feature extraction, computer vision, etc. Keywords : Image, Image Mining, Feature Extraction, Image Retrieval CBIR.
This document provides a review of different techniques for image retrieval from large databases, including text-based image retrieval and content-based image retrieval (CBIR). CBIR uses visual features extracted from images like color, texture, and shape to search for similar images. The document discusses some limitations of CBIR and proposes video-based image retrieval as a new direction. It also surveys recent research in areas like feature extraction, indexing, and discusses future directions like reducing the semantic gap between low-level features and high-level meanings.
This document discusses optimizing content-based image retrieval in peer-to-peer systems. It summarizes previous work on content-based image retrieval using multi-instance queries in peer-to-peer networks. The authors propose two optimizations to previous work: 1) clustering peers to reduce search time, and 2) constructing a search index at cluster heads to avoid searching each peer. Their experiments show the proposed approach reduces search time compared to previous work, with some reduction in accuracy that improves as more nodes are added. The authors plan future work to analyze performance with different cluster sizes and representations for faster search.
ATTENTION BASED IMAGE CAPTIONING USING DEEP LEARNINGNathan Mathis
The document describes a study on attention-based image captioning using deep learning. The study aims to generate image captions using an encoder-decoder model with an attention mechanism. The encoder is Google InceptionV3 which extracts image features, and the decoder is a GRU that generates captions. The model is trained on the MS COCO dataset and evaluated using BLEU score. Results show the attention mechanism helps focus on relevant image areas to produce descriptive captions.
META-HEURISTICS BASED ARF OPTIMIZATION FOR IMAGE RETRIEVALIJCSEIT Journal
The document proposes an approach combining automatic relevance feedback and particle swarm optimization for image retrieval. It constructs a visual feature database from image features like color moments and Gabor filters. For a query image, it retrieves similar images and generates automatic relevance feedback by labeling images as relevant or irrelevant. It then uses particle swarm optimization to re-weight features and retrieve more relevant images over multiple iterations, splitting the swarm in later iterations. An experiment on Corel images over 5 classes showed the approach could effectively retrieve relevant images through this meta-heuristic process without human interaction.
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.
An Enhance Image Retrieval of User Interest Using Query Specific Approach and...IJSRD
In recent years, image retrieval process has increased artistically. An image retrieval system is a process for searching and retrieving images from large amount of the image dataset. Color, texture and edge have been the primitive low level image descriptors in content based image retrieval systems. In this paper we discover a system which splits the search process into two stages. In the query specify approach the feature descriptors of a query image we re-extracted and then used to check the similarity between the query image and those images which is in database. In the evolution stage, the most relevant images where retrieved by using the Interactive genetic algorithm. IGA help the users to retrieve the images that are most relevant to the users’ need and SVM will rank the image as their title and as par time of search. So that user can get search image as par their requirements.
Image compression and reconstruction using a new approach by artificial neura...Hưng Đặng
This document describes a neural network approach to image compression and reconstruction. It discusses using a backpropagation neural network with three layers (input, hidden, output) to compress an image by representing it with fewer hidden units than input units, then reconstructing the image from the hidden unit values. It also covers preprocessing steps like converting images to YCbCr color space, downsampling chrominance, normalizing pixel values, and segmenting images into blocks for the neural network. The neural network weights are initially randomized and then trained using backpropagation to learn the image compression.
Image compression and reconstruction using a new approach by artificial neura...Hưng Đặng
This document describes a neural network approach to image compression and reconstruction. It discusses using a backpropagation neural network with three layers (input, hidden, output) to compress an image by representing it with fewer hidden units than input units, then reconstructing the image from the hidden unit values. It also covers preprocessing steps like converting images to YCbCr color space, downsampling chrominance, normalizing pixel values, and segmenting images into blocks for the neural network. The neural network weights are initially randomized and then trained using backpropagation to learn the image compression.
CNN FEATURES ARE ALSO GREAT AT UNSUPERVISED CLASSIFICATION cscpconf
This paper aims at providing insight on the transferability of deep CNN features to
unsupervised problems. We study the impact of different pretrained CNN feature extractors on
the problem of image set clustering for object classification as well as fine-grained
classification. We propose a rather straightforward pipeline combining deep-feature extraction
using a CNN pretrained on ImageNet and a classic clustering algorithm to classify sets of
images. This approach is compared to state-of-the-art algorithms in image-clustering and
provides better results. These results strengthen the belief that supervised training of deep CNN
on large datasets, with a large variability of classes, extracts better features than most carefully
designed engineering approaches, even for unsupervised tasks. We also validate our approach
on a robotic application, consisting in sorting and storing objects smartly based on clustering
Image Processing Compression and Reconstruction by Using New Approach Artific...CSCJournals
In this paper a neural network based image compression method is presented. Neural networks offer the potential for providing a novel solution to the problem of data compression by its ability to generate an internal data representation. This network, which is an application of back propagation network, accepts a large amount of image data, compresses it for storage or transmission, and subsequently restores it when desired. A new approach for reducing training time by reconstructing representative vectors has also been proposed. Performance of the network has been evaluated using some standard real world images. It is shown that the development architecture and training algorithm provide high compression ratio and low distortion while maintaining the ability to generalize and is very robust as well.
The complexity of landscape pattern mining is well stated due to its non-linear spatial image formation and
inhomogeneity of the satellite images. Land Ex tool of the literature work needs several seconds to answer input
image pattern query. The time duration of content based image retrieval depends on input query complexity. This
paper focuses on designing and implementing a training dataset to train NML (Neural network based Machine
Learning) algorithm to reduce the search time to improve the result accuracy. The performance evolution of
proposed NML CBIR (Content Based Image Retrieval) method will be used for comparison of satellite and natural
images by means of increasing speed and accuracy.
Keywords: Spatial Image, Satellite image, NML, CBIR
Research Inventy : International Journal of Engineering and Science is publis...researchinventy
This document summarizes a research paper that proposes a novel approach for content-based image retrieval using wavelet transform and hierarchical neural networks. The paper describes how wavelet transforms are used to extract features from images, and a neural network is trained on these features to classify and retrieve similar images. The system was tested on a database of 450 images across different categories. Initial results found an accuracy of about 70% when querying images. The paper concludes that while initial results are promising, further research is needed to explore different wavelet functions, feature extraction techniques, and classification methods to improve accuracy.
Because of the rapid growth in technology breakthroughs, including
multimedia and cell phones, Telugu character recognition (TCR) has recently
become a popular study area. It is still necessary to construct automated and
intelligent online TCR models, even if many studies have focused on offline
TCR models. The Telugu character dataset construction and validation using
an Inception and ResNet-based model are presented. The collection of 645
letters in the dataset includes 18 Achus, 38 Hallus, 35 Othulu, 34×16
Guninthamulu, and 10 Ankelu. The proposed technique aims to efficiently
recognize and identify distinctive Telugu characters online. This model's main
pre-processing steps to achieve its goals include normalization, smoothing,
and interpolation. Improved recognition performance can be attained by using
stochastic gradient descent (SGD) to optimize the model's hyperparameters.
Scientific workload execution on a distributed computing platform such as a
cloud environment is time-consuming and expensive. The scientific workload
has task dependencies with different service level agreement (SLA)
prerequisites at different levels. Existing workload scheduling (WS) designs
are not efficient in assuring SLA at the task level. Alongside, induces higher
costs as the majority of scheduling mechanisms reduce either time or energy.
In reducing, cost both energy and makespan must be optimized together for
allocating resources. No prior work has considered optimizing energy and
processing time together in meeting task level SLA requirements. This paper
presents task level energy and performance assurance-workload scheduling
(TLEPA-WS) algorithm for the distributed computing environment. The
TLEPA-WS guarantees energy minimization with the performance
requirement of the parallel application under a distributed computational
environment. Experiment results show a significant reduction in using energy
and makespan; thereby reducing the cost of workload execution in comparison
with various standard workload execution models.
More Related Content
Similar to Efficient content-based image retrieval using integrated dual deep convolutional neural network
A deep locality-sensitive hashing approach for achieving optimal image retri...IJECEIAES
The proposed method achieves optimal image retrieval through a deep locality-sensitive hashing approach. It extracts both low-level visual features and high-level semantic features from different layers of a CNN model. Locality-sensitive hashing is applied to the features to generate hash codes, which are then used to quickly retrieve similar images based on Hamming distance. Experimental results on CIFAR-10 and NUS-WIDE datasets show the proposed method outperforms other hash-based image retrieval methods in terms of accuracy and retrieval time.
End-to-end deep auto-encoder for segmenting a moving object with limited tra...IJECEIAES
The document proposes two end-to-end deep auto-encoder approaches for segmenting moving objects from surveillance videos when limited training data is available. The first approach uses transfer learning with a pre-trained VGG-16 model as the encoder and its transposed architecture as the decoder. The second approach uses a multi-depth auto-encoder with convolutional and upsampling layers. Both approaches apply data augmentation techniques like PCA and traditional methods to increase the training data size. The models are trained and evaluated on the CDnet2014 dataset, achieving better performance than other models trained with limited data.
The advents in this technological era have resulted into enormous pool of information. This information is
stored at multiple places globally, in multiple formats. This article highlights a methodology for extracting
the video lectures delivered by experts in the domain of Computer Science by using Generalized Gamma
Mixture Model. The feature extraction is based on the DCT transformations. In order to propose the model,
the data set is pooled from the YouTube video lectures in the domain of Computer Science. The outputs
generated are evaluated using Precision and Recall.
Precision face image retrieval by extracting the face features and comparing ...prjpublications
This document describes a proposed method for improving content-based face image retrieval. The method uses two orthogonal techniques: attribute-enhanced sparse coding and attribute-embedded inverted indexing. Attribute-enhanced sparse coding exploits global features to construct semantic codewords offline. Attribute-embedded inverted indexing considers local query image features in a binary signature to efficiently retrieve images. By combining these techniques, the method reduces errors and achieves better face image extraction from databases compared to existing content-based retrieval systems. It works by extracting features from the query image, matching them to database images, and returning ranked results.
Content-based Image Retrieval System for an Image Gallery Search Application IJECEIAES
Content-based image retrieval is a process framework that applies computer vision techniques for searching and managing large image collections more efficiently. With the growth of large digital image collections triggered by rapid advances in electronic storage capacity and computing power, there is a growing need for devices and computer systems to support efficient browsing, searching, and retrieval for image collections. Hence, the aim of this project is to develop a content-based image retrieval system that can be implemented in an image gallery desktop application to allow efficient browsing through three different search modes: retrieval by image query, retrieval by facial recognition, and retrieval by text or tags. In this project, the MPEG-7-like Powered Localized Color and Edge Directivity Descriptor is used to extract the feature vectors of the image database and the facial recognition system is built around the Eigenfaces concept. A graphical user interface with the basic functionality of an image gallery application is also developed to implement the three search modes. Results show that the application is able to retrieve and display images in a collection as thumbnail previews with high retrieval accuracy and medium relevance and the computational requirements for subsequent searches were significantly reduced through the incorporation of text-based image retrieval as one of the search modes. All in all, this study introduces a simple and convenient way of offline image searches on desktop computers and provides a stepping stone to future content-based image retrieval systems built for similar purposes.
RECURRENT FEATURE GROUPING AND CLASSIFICATION MODEL FOR ACTION MODEL PREDICTI...IJDKP
Content based retrieval has an advantage of higher prediction accuracy as compared to tagging based approach. However, the complexity in its representation and classification approach, results in lower processing accuracy and computation overhead. The correlative nature of the feature data are un-explored in the conventional modeling, where all the data features are taken as a set of feature values to give a decision. The recurrent feature class attribute is observed for the feature regrouping in action model prediction. In this paper a co-relative information, bounding grouping approach is suggested for action model prediction in CBMR application. The co-relative recurrent feature mapping results in faster retrieval process as compared to the conventional retrieval system.
RECURRENT FEATURE GROUPING AND CLASSIFICATION MODEL FOR ACTION MODEL PREDICTI...IJDKP
Content based retrieval has an advantage of higher prediction accuracy as compared to tagging based approach. However, the complexity in its representation and classification approach, results in lower processing accuracy and computation overhead. The correlative nature of the feature data are un-explored in the conventional modeling, where all the data features are taken as a set of feature values to give a decision. The recurrent feature class attribute is observed for the feature regrouping in action model prediction. In this paper a co-relative information, bounding grouping approach is suggested for action model prediction in CBMR application. The co-relative recurrent feature mapping results in faster retrieval process as compared to the conventional retrieval system.
Mining of images using retrieval techniqueseSAT Journals
Abstract Today’s world is digital. Use of different social websites is become a part of our day to day life. These social websites have acquired great popularity because of their user-friendly features and their content. Nowadays internet and mobile networks are widely used everywhere and so use of images. Because of evolution in hardware as well as software it is possible to store large amount of multimedia data. In short, now a day it becomes easy to store huge amount of images by using image processing techniques. As number of images and databases are increasing day by day, there is a need for new image retrieval techniques that should be fulfilled. Image mining is derived from data mining which is a method to extract information from digital image. The purpose of this paper is to present a review of the various image mining techniques used in different applications as image retrieval, Matching, Pattern recognition etc.given by different researchers. An information or knowledge can be extracted from the image by using image mining technique. This paper proposes the survey in image retrieval techniques. Image retrieval and data mining have huge applications in the sector of image processing, pattern recognition, image matching, image mining, feature extraction, computer vision, etc. Keywords : Image, Image Mining, Feature Extraction, Image Retrieval CBIR.
This document provides a review of different techniques for image retrieval from large databases, including text-based image retrieval and content-based image retrieval (CBIR). CBIR uses visual features extracted from images like color, texture, and shape to search for similar images. The document discusses some limitations of CBIR and proposes video-based image retrieval as a new direction. It also surveys recent research in areas like feature extraction, indexing, and discusses future directions like reducing the semantic gap between low-level features and high-level meanings.
This document discusses optimizing content-based image retrieval in peer-to-peer systems. It summarizes previous work on content-based image retrieval using multi-instance queries in peer-to-peer networks. The authors propose two optimizations to previous work: 1) clustering peers to reduce search time, and 2) constructing a search index at cluster heads to avoid searching each peer. Their experiments show the proposed approach reduces search time compared to previous work, with some reduction in accuracy that improves as more nodes are added. The authors plan future work to analyze performance with different cluster sizes and representations for faster search.
ATTENTION BASED IMAGE CAPTIONING USING DEEP LEARNINGNathan Mathis
The document describes a study on attention-based image captioning using deep learning. The study aims to generate image captions using an encoder-decoder model with an attention mechanism. The encoder is Google InceptionV3 which extracts image features, and the decoder is a GRU that generates captions. The model is trained on the MS COCO dataset and evaluated using BLEU score. Results show the attention mechanism helps focus on relevant image areas to produce descriptive captions.
META-HEURISTICS BASED ARF OPTIMIZATION FOR IMAGE RETRIEVALIJCSEIT Journal
The document proposes an approach combining automatic relevance feedback and particle swarm optimization for image retrieval. It constructs a visual feature database from image features like color moments and Gabor filters. For a query image, it retrieves similar images and generates automatic relevance feedback by labeling images as relevant or irrelevant. It then uses particle swarm optimization to re-weight features and retrieve more relevant images over multiple iterations, splitting the swarm in later iterations. An experiment on Corel images over 5 classes showed the approach could effectively retrieve relevant images through this meta-heuristic process without human interaction.
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.
An Enhance Image Retrieval of User Interest Using Query Specific Approach and...IJSRD
In recent years, image retrieval process has increased artistically. An image retrieval system is a process for searching and retrieving images from large amount of the image dataset. Color, texture and edge have been the primitive low level image descriptors in content based image retrieval systems. In this paper we discover a system which splits the search process into two stages. In the query specify approach the feature descriptors of a query image we re-extracted and then used to check the similarity between the query image and those images which is in database. In the evolution stage, the most relevant images where retrieved by using the Interactive genetic algorithm. IGA help the users to retrieve the images that are most relevant to the users’ need and SVM will rank the image as their title and as par time of search. So that user can get search image as par their requirements.
Image compression and reconstruction using a new approach by artificial neura...Hưng Đặng
This document describes a neural network approach to image compression and reconstruction. It discusses using a backpropagation neural network with three layers (input, hidden, output) to compress an image by representing it with fewer hidden units than input units, then reconstructing the image from the hidden unit values. It also covers preprocessing steps like converting images to YCbCr color space, downsampling chrominance, normalizing pixel values, and segmenting images into blocks for the neural network. The neural network weights are initially randomized and then trained using backpropagation to learn the image compression.
Image compression and reconstruction using a new approach by artificial neura...Hưng Đặng
This document describes a neural network approach to image compression and reconstruction. It discusses using a backpropagation neural network with three layers (input, hidden, output) to compress an image by representing it with fewer hidden units than input units, then reconstructing the image from the hidden unit values. It also covers preprocessing steps like converting images to YCbCr color space, downsampling chrominance, normalizing pixel values, and segmenting images into blocks for the neural network. The neural network weights are initially randomized and then trained using backpropagation to learn the image compression.
CNN FEATURES ARE ALSO GREAT AT UNSUPERVISED CLASSIFICATION cscpconf
This paper aims at providing insight on the transferability of deep CNN features to
unsupervised problems. We study the impact of different pretrained CNN feature extractors on
the problem of image set clustering for object classification as well as fine-grained
classification. We propose a rather straightforward pipeline combining deep-feature extraction
using a CNN pretrained on ImageNet and a classic clustering algorithm to classify sets of
images. This approach is compared to state-of-the-art algorithms in image-clustering and
provides better results. These results strengthen the belief that supervised training of deep CNN
on large datasets, with a large variability of classes, extracts better features than most carefully
designed engineering approaches, even for unsupervised tasks. We also validate our approach
on a robotic application, consisting in sorting and storing objects smartly based on clustering
Image Processing Compression and Reconstruction by Using New Approach Artific...CSCJournals
In this paper a neural network based image compression method is presented. Neural networks offer the potential for providing a novel solution to the problem of data compression by its ability to generate an internal data representation. This network, which is an application of back propagation network, accepts a large amount of image data, compresses it for storage or transmission, and subsequently restores it when desired. A new approach for reducing training time by reconstructing representative vectors has also been proposed. Performance of the network has been evaluated using some standard real world images. It is shown that the development architecture and training algorithm provide high compression ratio and low distortion while maintaining the ability to generalize and is very robust as well.
The complexity of landscape pattern mining is well stated due to its non-linear spatial image formation and
inhomogeneity of the satellite images. Land Ex tool of the literature work needs several seconds to answer input
image pattern query. The time duration of content based image retrieval depends on input query complexity. This
paper focuses on designing and implementing a training dataset to train NML (Neural network based Machine
Learning) algorithm to reduce the search time to improve the result accuracy. The performance evolution of
proposed NML CBIR (Content Based Image Retrieval) method will be used for comparison of satellite and natural
images by means of increasing speed and accuracy.
Keywords: Spatial Image, Satellite image, NML, CBIR
Research Inventy : International Journal of Engineering and Science is publis...researchinventy
This document summarizes a research paper that proposes a novel approach for content-based image retrieval using wavelet transform and hierarchical neural networks. The paper describes how wavelet transforms are used to extract features from images, and a neural network is trained on these features to classify and retrieve similar images. The system was tested on a database of 450 images across different categories. Initial results found an accuracy of about 70% when querying images. The paper concludes that while initial results are promising, further research is needed to explore different wavelet functions, feature extraction techniques, and classification methods to improve accuracy.
Similar to Efficient content-based image retrieval using integrated dual deep convolutional neural network (20)
Because of the rapid growth in technology breakthroughs, including
multimedia and cell phones, Telugu character recognition (TCR) has recently
become a popular study area. It is still necessary to construct automated and
intelligent online TCR models, even if many studies have focused on offline
TCR models. The Telugu character dataset construction and validation using
an Inception and ResNet-based model are presented. The collection of 645
letters in the dataset includes 18 Achus, 38 Hallus, 35 Othulu, 34×16
Guninthamulu, and 10 Ankelu. The proposed technique aims to efficiently
recognize and identify distinctive Telugu characters online. This model's main
pre-processing steps to achieve its goals include normalization, smoothing,
and interpolation. Improved recognition performance can be attained by using
stochastic gradient descent (SGD) to optimize the model's hyperparameters.
Scientific workload execution on a distributed computing platform such as a
cloud environment is time-consuming and expensive. The scientific workload
has task dependencies with different service level agreement (SLA)
prerequisites at different levels. Existing workload scheduling (WS) designs
are not efficient in assuring SLA at the task level. Alongside, induces higher
costs as the majority of scheduling mechanisms reduce either time or energy.
In reducing, cost both energy and makespan must be optimized together for
allocating resources. No prior work has considered optimizing energy and
processing time together in meeting task level SLA requirements. This paper
presents task level energy and performance assurance-workload scheduling
(TLEPA-WS) algorithm for the distributed computing environment. The
TLEPA-WS guarantees energy minimization with the performance
requirement of the parallel application under a distributed computational
environment. Experiment results show a significant reduction in using energy
and makespan; thereby reducing the cost of workload execution in comparison
with various standard workload execution models.
Investigating human subjects is the goal of predicting human emotions in the
real world scenario. A significant number of psychological effects require
(feelings) to be produced, directly releasing human emotions. The
development of effect theory leads one to believe that one must be aware of
one's sentiments and emotions to forecast one's behavior. The proposed line
of inquiry focuses on developing a reliable model incorporating
neurophysiological data into actual feelings. Any change in emotional affect
will directly elicit a response in the body's physiological systems. This
approach is named after the notion of Gaussian mixture models (GMM). The
statistical reaction following data processing, quantitative findings on emotion
labels, and coincidental responses with training samples all directly impact the
outcomes that are accomplished. In terms of statistical parameters such as
population mean and standard deviation, the suggested method is evaluated
compared to a technique considered to be state-of-the-art. The proposed
system determines an individual's emotional state after a minimum of 6
iterative learning using the Gaussian expectation-maximization (GEM)
statistical model, in which the iterations tend to continue to zero error. Perhaps
each of these improves predictions while simultaneously increasing the
amount of value extracted.
Early diagnosis of cancers is a major requirement for patients and a
complicated job for the oncologist. If it is diagnosed early, it could have made
the patient more likely to live. For a few decades, fuzzy logic emerged as an
emphatic technique in the identification of diseases like different types of
cancers. The recognition of cancer diseases mostly operated with inexactness,
inaccuracy, and vagueness. This paper aims to design the fuzzy expert system
(FES) and its implementation for the detection of prostate cancer. Specifically,
prostate-specific antigen (PSA), prostate volume (PV), age, and percentage
free PSA (%FPSA) are used to determine prostate cancer risk (PCR), while
PCR serves as an output parameter. Mamdani fuzzy inference method is used
to calculate a range of PCR. The system provides a scale of risk of prostate
cancer and clears the path for the oncologist to determine whether their
patients need a biopsy. This system is fast as it requires minimum calculation
and hence comparatively less time which reduces mortality and morbidity and
is more reliable than other economic systems and can be frequently used by
doctors.
The biomedical profession has gained importance due to the rapid and accurate diagnosis of clinical patients using computer-aided diagnosis (CAD) tools.
The diagnosis and treatment of Alzheimer’s disease (AD) using complementary multimodalities can improve the quality of life and mental state of patients.
In this study, we integrated a lightweight custom convolutional neural network
(CNN) model and nature-inspired optimization techniques to enhance the performance, robustness, and stability of progress detection in AD. A multi-modal
fusion database approach was implemented, including positron emission tomography (PET) and magnetic resonance imaging (MRI) datasets, to create a fused
database. We compared the performance of custom and pre-trained deep learning models with and without optimization and found that employing natureinspired algorithms like the particle swarm optimization algorithm (PSO) algorithm significantly improved system performance. The proposed methodology,
which includes a fused multimodality database and optimization strategy, improved performance metrics such as training, validation, test accuracy, precision, and recall. Furthermore, PSO was found to improve the performance of
pre-trained models by 3-5% and custom models by up to 22%. Combining different medical imaging modalities improved the overall model performance by
2-5%. In conclusion, a customized lightweight CNN model and nature-inspired
optimization techniques can significantly enhance progress detection, leading to
better biomedical research and patient care.
Class imbalance is a pervasive issue in the field of disease classification from
medical images. It is necessary to balance out the class distribution while training a model. However, in the case of rare medical diseases, images from affected
patients are much harder to come by compared to images from non-affected
patients, resulting in unwanted class imbalance. Various processes of tackling
class imbalance issues have been explored so far, each having its fair share of
drawbacks. In this research, we propose an outlier detection based image classification technique which can handle even the most extreme case of class imbalance. We have utilized a dataset of malaria parasitized and uninfected cells. An
autoencoder model titled AnoMalNet is trained with only the uninfected cell images at the beginning and then used to classify both the affected and non-affected
cell images by thresholding a loss value. We have achieved an accuracy, precision, recall, and F1 score of 98.49%, 97.07%, 100%, and 98.52% respectively,
performing better than large deep learning models and other published works.
As our proposed approach can provide competitive results without needing the
disease-positive samples during training, it should prove to be useful in binary
disease classification on imbalanced datasets.
Recently, plant identification has become an active trend due to encouraging
results achieved in plant species detection and plant classification fields
among numerous available plants using deep learning methods. Therefore,
plant classification analysis is performed in this work to address the problem
of accurate plant species detection in the presence of multiple leaves together,
flowers, and noise. Thus, a convolutional neural network based deep feature
learning and classification (CNN-DFLC) model is designed to analyze
patterns of plant leaves and perform classification using generated finegrained feature weights. The proposed CNN-DFLC model precisely estimates
which the given image belongs to which plant species. Several layers and
blocks are utilized to design the proposed CNN-DFLC model. Fine-grained
feature weights are obtained using convolutional and pooling layers. The
obtained feature maps in training are utilized to predict labels and model
performance is tested on the Vietnam plant image (VPN-200) dataset. This
dataset consists of a total number of 20,000 images and testing results are
achieved in terms of classification accuracy, precision, recall, and other
performance metrics. The mean classification accuracy obtained using the
proposed CNN-DFLC model is 96.42% considering all 200 classes from the
VPN-200 dataset.
Big data as a service (BDaaS) platform is widely used by various
organizations for handling and processing the high volume of data generated
from different internet of things (IoT) devices. Data generated from these IoT
devices are kept in the form of big data with the help of cloud computing
technology. Researchers are putting efforts into providing a more secure and
protected access environment for the data available on the cloud. In order to
create a safe, distributed, and decentralised environment in the cloud,
blockchain technology has emerged as a useful tool. In this research paper, we
have proposed a system that uses blockchain technology as a tool to regulate
data access that is provided by BDaaS platforms. We are securing the access
policy of data by using a modified form of ciphertext policy-attribute based
encryption (CP-ABE) technique with the help of blockchain technology. For
secure data access in BDaaS, algorithms have been created using a mix of CPABE with blockchain technology. Proposed smart contract algorithms are
implemented using Eclipse 7.0 IDE and the cloud environment has been
simulated on CloudSim tool. Results of key generation time, encryption time,
and decryption time has been calculated and compared with access control
mechanism without blockchain technology.
Internet of things (IoT) has become one of the eminent phenomena in human
life along with its collaboration with wireless sensor networks (WSNs), due
to enormous growth in the domain; there has been a demand to address the
various issues regarding it such as energy consumption, redundancy, and
overhead. Data aggregation (DA) is considered as the basic mechanism to
minimize the energy efficiency and communication overhead; however,
security plays an important role where node security is essential due to the
volatile nature of WSN. Thus, we design and develop proximate node aware
secure data aggregation (PNA-SDA). In the PNA-SDA mechanism, additional
data is used to secure the original data, and further information is shared with
the proximate node; moreover, further security is achieved by updating the
state each time. Moreover, the node that does not have updated information is
considered as the compromised node and discarded. PNA-SDA is evaluated
considering the different parameters like average energy consumption, and
average deceased node; also, comparative analysis is carried out with the
existing model in terms of throughput and correct packet identification.
Drones provide an alternative progression in protection submissions since
they are capable of conducting autonomous seismic investigations. Recent
advancement in unmanned aerial vehicle (UAV) communication is an internet
of a drone combined with 5G networks. Because of the quick utilization of
rapidly progressed registering frameworks besides 5G officialdoms, the
information from the user is consistently refreshed and pooled. Thus, safety
or confidentiality is vital among clients, and a proficient substantiation
methodology utilizing a vigorous sanctuary key. Conventional procedures
ensure a few restrictions however taking care of the assault arrangements in
information transmission over the internet of drones (IOD) environmental
frameworks. A unique hyperelliptical curve (HEC) cryptographically based
validation system is proposed to provide protected data facilities among
drones. The proposed method has been compared with the existing methods
in terms of packet loss rate, computational cost, and delay and thereby
provides better insight into efficient and secure communication. Finally, the
simulation results show that our strategy is efficient in both computation and
communication.
Monitoring behavior, numerous actions, or any such information is considered
as surveillance and is done for information gathering, influencing, managing,
or directing purposes. Citizens employ surveillance to safeguard their
communities. Governments do this for the purposes of intelligence collection,
including espionage, crime prevention, the defense of a method, a person, a
group, or an item; or the investigation of criminal activity. Using an internet
of things (IoT) rover, the area will be secured with better secrecy and
efficiency instead of humans, will provide an additional safety step. In this
paper, there is a discussion about an IoT rover for remote surveillance based
around a Raspberry Pi microprocessor which will be able to monitor a
closed/open space. This rover will allow safer survey operations and would
help to reduce the risks involved with it.
In a world where climate change looms large the spotlight often shines on
greenhouse gases, but the shadow of man-made aerosols should not be
underestimated. These tiny particles play a pivotal role in disrupting Earth's
radiative equilibrium, yet many mysteries surround their influence on various
physical aspects of our planet. The root of these mysteries lies in the limited
data we have on aerosol sources, formation processes, conversion dynamics,
and collection methods. Aerosols, composed of particulate matter (PM),
sulfates, and nitrates, hold significant sway across the hemisphere. Accurate
measurement demands the refinement of in-situ, satellite, and ground-based
techniques. As aerosols interact intricately with the environment, their full
impact remains an enigma. Enter a groundbreaking study in Morocco that
dared to compare an internet of thing (IoT) system with satellite-based
atmospheric models, with a focus on fine particles below 10 and 2.5
micrometers in diameter. The initial results, particularly in regions abundant
with extraction pits, shed light on the IoT system's potential to decode
aerosols' role in the grand narrative of climate change. These findings inspire
hope as we confront the formidable global challenge of climate change.
The use of technology has a significant impact to reduce the consequences of
accidents. Sensors, small components that detect interactions experienced by
various components, play a crucial role in this regard. This study focuses on
how the MPU6050 sensor module can be used to detect the movement of
people who are falling, defined as the inability of the lower body, including
the hips and feet, to support the body effectively. An airbag system is
proposed to reduce the impact of a fall. The data processing method in this
study involves the use of a threshold value to identify falling motion. The
results of the study have identified a threshold value for falling motion,
including an acceleration relative (AR) value of less than or equal to 0.38 g,
an angle slope of more than or equal to 40 degrees, and an angular velocity
of more than or equal to 30 °/s. The airbag system is designed to inflate
faster than the time of impact, with a gas flow rate of 0.04876 m3
/s and an
inflating time of 0.05 s. The overall system has a specificity value of 100%,
a sensitivity of 85%, and an accuracy of 94%.
The fundamental principle of the paper is that the soil moisture sensor obtains
the moisture content level of the soil sample. The water pump is automatically
activated if the moisture content is insufficient, which causes water to flow
into the soil. The water pump is immediately turned off when the moisture
content is high enough. Smart home, smart city, smart transportation, and
smart farming are just a few of the new intelligent ideas that internet of things
(IoT) includes. The goal of this method is to increase productivity and
decrease manual labour among farmers. In this paper, we present a system for
monitoring and regulating water flow that employs a soil moisture sensor to
keep track of soil moisture content as well as the land’s water level to keep
track of and regulate the amount of water supplied to the plant. The device
also includes an automated led lighting system.
In order to provide sensing services to low-powered IoT devices, wireless sensor networks (WSNs) organize specialized transducers into networks. Energy usage is one of the most important design concerns in WSN because it is very hard to replace or recharge the batteries in sensor nodes. For an energy-constrained network, the clustering technique is crucial in preserving battery life. By strategically selecting a cluster head (CH), a network's load can be balanced, resulting in decreased energy usage and extended system life. Although clustering has been predominantly used in the literature, the concept of chain-based clustering has not yet been explored. As a result, in this paper, we employ a chain-based clustering architecture for data dissemination in the network. Furthermore, for CH selection, we employ the coati optimisation algorithm, which was recently proposed and has demonstrated significant improvement over other optimization algorithms. In this method, the parameters considered for selecting the CH are energy, node density, distance, and the network’s average energy. The simulation results show tremendous improvement over the competitive cluster-based routing algorithms in the context of network lifetime, stability period (first node dead), transmission rate, and the network's power reserves.
The construction industry is an industry that is always surrounded by
uncertainties and risks. The industry is always associated with a threatindustry which has a complex, tedious layout and techniques characterized by
unpredictable circumstances. It comprises a variety of human talents and the
coordination of different areas and activities associated with it. In this
competitive era of the construction industry, delays and cost overruns of the
project are often common in every project and the causes of that are also
common. One of the problems which we are trying to cater to is the improper
handling of materials at the construction site. In this paper, we propose
developing a system that is capable of tracking construction material on site
that would benefit the contractor and client for better control over inventory
on-site and to minimize loss of material that occurs due to theft and misplacing
of materials.
Today, health monitoring relies heavily on technological advancements. This
study proposes a low-power wide-area network (LPWAN) based, multinodal
health monitoring system to monitor vital physiological data. The suggested
system consists of two nodes, an indoor node, and an outdoor node, and the
nodes communicate via long range (LoRa) transceivers. Outdoor nodes use an
MPU6050 module, heart rate, oxygen pulse, temperature, and skin resistance
sensors and transmit sensed values to the indoor node. We transferred the data
received by the master node to the cloud using the Adafruit cloud service. The
system can operate with a coverage of 4.5 km, where the optimal distance
between outdoor sensor nodes and the indoor master node is 4 km. To further
predict fall detection, various machine learning classification techniques have
been applied. Upon comparing various classifier techniques, the decision tree
method achieved an accuracy of 0.99864 with a training and testing ratio of
70:30. By developing accurate prediction models, we can identify high-risk
individuals and implement preventative measures to reduce the likelihood of
a fall occurring. Remote monitoring of the health and physical status of elderly
people has proven to be the most beneficial application of this technology.
The effectiveness of adaptive filters are mainly dependent on the design
techniques and the algorithm of adaptation. The most common adaptation
technique used is least mean square (LMS) due its computational simplicity.
The application depends on the adaptive filter configuration used and are well
known for system identification and real time applications. In this work, a
modified delayed μ-law proportionate normalized least mean square
(DMPNLMS) algorithm has been proposed. It is the improvised version of the
µ-law proportionate normalized least mean square (MPNLMS) algorithm.
The algorithm is realized using Ladner-Fischer type of parallel prefix
logarithmic adder to reduce the silicon area. The simulation and
implementation of very large-scale integration (VLSI) architecture are done
using MATLAB, Vivado suite and complementary metal–oxide–
semiconductor (CMOS) 90 nm technology node using Cadence RTL and
Genus Compiler respectively. The DMPNLMS method exhibits a reduction
in mean square error, a higher rate of convergence, and more stability. The
synthesis results demonstrate that it is area and delay effective, making it
practical for applications where a faster operating speed is required.
The increasing demand for faster, robust, and efficient device development of enabling technology to mass production of industrial research in circuit design deals with challenges like size, efficiency, power, and scalability. This paper, presents a design and analysis of low power high speed full adder using negative capacitance field effecting transistors. A comprehensive study is performed with adiabatic logic and reversable logic. The performance of full adder is studied with metal oxide field effect transistor (MOSFET) and negative capacitance field effecting (NCFET). The NCFET based full adder offers a low power and high speed compared with conventional MOSFET. The complete design and analysis are performed using cadence virtuoso. The adiabatic logic offering low delay of 0.023 ns and reversable logic is offering low power of 7.19 mw.
The global agriculture system faces significant challenges in meeting the
growing demand for food production, particularly given projections that the
world's population will reach 70% by 2050. Hydroponic farming is an
increasingly popular technique in this field, offering a promising solution to
these challenges. This paper will present the improvement of the current
traditional hydroponic method by providing a system that can be used to
monitor and control the important element in order to help the plant grow up
smoothly. This proposed system is quite efficient and user-friendly that can
be used by anyone. This is a combination of a traditional hydroponic system,
an automatic control system and a smartphone. The primary objective is to
develop a smart system capable of monitoring and controlling potential
hydrogen (pH) levels, a key factor that affects hydroponic plant growth.
Ultimately, this paper offers an alternative approach to address the challenges
of the existing agricultural system and promote the production of clean,
disease-free, and healthy food for a better future.
More from International Journal of Reconfigurable and Embedded Systems (20)
KuberTENes Birthday Bash Guadalajara - K8sGPT first impressionsVictor Morales
K8sGPT is a tool that analyzes and diagnoses Kubernetes clusters. This presentation was used to share the requirements and dependencies to deploy K8sGPT in a local environment.
Electric vehicle and photovoltaic advanced roles in enhancing the financial p...IJECEIAES
Climate change's impact on the planet forced the United Nations and governments to promote green energies and electric transportation. The deployments of photovoltaic (PV) and electric vehicle (EV) systems gained stronger momentum due to their numerous advantages over fossil fuel types. The advantages go beyond sustainability to reach financial support and stability. The work in this paper introduces the hybrid system between PV and EV to support industrial and commercial plants. This paper covers the theoretical framework of the proposed hybrid system including the required equation to complete the cost analysis when PV and EV are present. In addition, the proposed design diagram which sets the priorities and requirements of the system is presented. The proposed approach allows setup to advance their power stability, especially during power outages. The presented information supports researchers and plant owners to complete the necessary analysis while promoting the deployment of clean energy. The result of a case study that represents a dairy milk farmer supports the theoretical works and highlights its advanced benefits to existing plants. The short return on investment of the proposed approach supports the paper's novelty approach for the sustainable electrical system. In addition, the proposed system allows for an isolated power setup without the need for a transmission line which enhances the safety of the electrical network
Harnessing WebAssembly for Real-time Stateless Streaming PipelinesChristina Lin
Traditionally, dealing with real-time data pipelines has involved significant overhead, even for straightforward tasks like data transformation or masking. However, in this talk, we’ll venture into the dynamic realm of WebAssembly (WASM) and discover how it can revolutionize the creation of stateless streaming pipelines within a Kafka (Redpanda) broker. These pipelines are adept at managing low-latency, high-data-volume scenarios.
A review on techniques and modelling methodologies used for checking electrom...nooriasukmaningtyas
The proper function of the integrated circuit (IC) in an inhibiting electromagnetic environment has always been a serious concern throughout the decades of revolution in the world of electronics, from disjunct devices to today’s integrated circuit technology, where billions of transistors are combined on a single chip. The automotive industry and smart vehicles in particular, are confronting design issues such as being prone to electromagnetic interference (EMI). Electronic control devices calculate incorrect outputs because of EMI and sensors give misleading values which can prove fatal in case of automotives. In this paper, the authors have non exhaustively tried to review research work concerned with the investigation of EMI in ICs and prediction of this EMI using various modelling methodologies and measurement setups.
6th International Conference on Machine Learning & Applications (CMLA 2024)ClaraZara1
6th International Conference on Machine Learning & Applications (CMLA 2024) will provide an excellent international forum for sharing knowledge and results in theory, methodology and applications of on Machine Learning & Applications.
Understanding Inductive Bias in Machine LearningSUTEJAS
This presentation explores the concept of inductive bias in machine learning. It explains how algorithms come with built-in assumptions and preferences that guide the learning process. You'll learn about the different types of inductive bias and how they can impact the performance and generalizability of machine learning models.
The presentation also covers the positive and negative aspects of inductive bias, along with strategies for mitigating potential drawbacks. We'll explore examples of how bias manifests in algorithms like neural networks and decision trees.
By understanding inductive bias, you can gain valuable insights into how machine learning models work and make informed decisions when building and deploying them.
Efficient content-based image retrieval using integrated dual deep convolutional neural network
1. International Journal of Reconfigurable and Embedded Systems (IJRES)
Vol. 12, No. 2, July 2023, pp. 297~304
ISSN: 2089-4864, DOI: 10.11591/ijres.v12.i2.pp297-304 297
Journal homepage: http://ijres.iaescore.com
Efficient content-based image retrieval using integrated dual
deep convolutional neural network
Feroza D. Mirajkar1
, Ruksar Fatima2
, Shaik A. Qadeer3
1
Department of Electronics and Communication Engineering, Khaja Banda Nawaz College of Engineering (KBNCE), Kalaburagi, India
2
HOD Computer Science and Engineering, Khaja Banda Nawaz College of Engineering (KBNCE), Kalaburagi, India
3
Department of Electrical and Electronics Engineering, Muffakham Jah College of Engineering and Technology, Hyderabad, India
Article Info ABSTRACT
Article history:
Received Jul 26, 2022
Revised Oct 15, 2022
Accepted Dec 10, 2022
Content-based image retrieval (CBIR) uses the content features for
retrieving and searching the images in a given large database. Earlier,
different hand feature descriptor designs are researched based on cues that
are visual such as shape, colour, and texture used to represent these images.
Although, deep learning technologies have widely been applied as an
alternative to designing engineering that is dominant for over a decade. The
features are automatically learnt through the data. This research work
proposes integrated dual deep convolutional neural network (IDD-CNN),
IDD-CNN comprises two distinctive CNN, first CNN exploits the features
and further custom CNN is designed for exploiting the custom features.
Moreover, a novel directed graph is designed that comprises the two blocks
i.e. learning block and memory block which helps in finding the similarity
among images; since this research considers the large dataset, an optimal
strategy is introduced for compact features. Moreover, IDD-CNN is
evaluated considering the two distinctive benchmark datasets the oxford
dataset considering mean average precision (mAP) metrics and comparative
analysis shows IDD-CNN outperforms the other existing model.
Keywords:
Content-based image retrieval
Convolutional neural network
Image retrieval
Images
Integrated dual deep-CNN
This is an open access article under the CC BY-SA license.
Corresponding Author:
Feroza D. Mirajkar
Department of Electronics and Communication Engineering
Khaja Bandanawaz College of engineering (KBNCE)
Kalaburagi, Karnataka, India
Email: mmferoza@gmail.com
1. INTRODUCTION
In the last two decades, there has been enormous development in technologies, which led to the huge
growth in internet usage, smartphone and digital cameras. Moreover, this phenomenon increases storing and
sharing of multimedia data such as images or videos. Image is one of the complex forms of data; hence,
searching for the relevant image from an archive is considered one of the challenging tasks [1]. In such an
approach, the user submits the query by entering some keywords or text that matches with text or keywords
in the archive. However, this process also retrieves the images, which are not relevant to the query [2], [3].
Content-based image retrieval (CBIR) is one of the tasks that are designed by defining the problem of
searching the image based on semantic matching given a large dataset.
CBIR approach aims at searching images based on their visual content. The query image is given
and the aim is to find the image that contains a similar scene or object, which might be captured under
various conditions [3]-[5]. Category level CBIR aims at finding the same image class as a defined query.
Figure 1 shows the general process of deep learning backed CBIR [6]. Six blocks are present in Figure 1, but
only two of them-the content information and the query images-are retrieved using a deep learning-based
2. ISSN: 2089-4864
Int J Reconfigurable & Embedded Syst, Vol. 12, No. 2, July 2023: 297-304
298
methodology. While one deep learning approach block processes the query image, another deep learning
approach block extracts the feature. Additionally, notable characteristics are chosen from the retrieved
feature. Additionally, these features are contrasted during feature matching, and the results are produced as
metrics [7]-[9].
Figure 1. Content-based image retrieval using the deep learning approach
Recently, this problem has been resolved by the use of convolutional neural networks (CNN), this
method also provides a higher rate of accuracy. The success of CNN has attained a load of attention towards
the technologies relating to neural networks considering tasks of image classification. The resulting success
is complete because of the enormous datasets annotated such as ImageNet. The training of data is an
expensive process with manual annotation is more prone to errors. The trained network for the classification
of images has good abilities for adaptation. The particular use of CNN activations was utilized for training
the classification tasks of image descriptors that are off-shelf as well as being adapted for various tasks has
resulted in success. Specifically considering image retrieval, different approaches have directly utilized the
network activations as features of images as well as performed the image searching successfully [8].
Powerful, major features have been successfully learnt using deep learning. Although, some major challenges
arise concerning: i) semantic gap reduction, ii) improvising scalability of retrieval, and iii) balance of
efficiency as well as the accuracy of retrieval.
An approach of order less fusion of multilayers (MOF) is proposed [10] that has been inspired by an
order less pooling of multilayers (MOP) [11] utilized for the retrieval of images. Although, the local features
have no discrete role in the differentiation of features that are subtle due to the local as well as global features
being treated as identical. Zhang et al. [12], a solution is introduced that emphasizes varying multiple
instances of the graph (VMIG) for which a constant semantic space is studied to save its query semantics that
are diverse. The retrieving task has been formulated with different instances of studying the problems for
connecting the diverse features to the modalities. Particularly, a vibrational auto encoder that is query guided
is used for modelling the constant semantic space rather than studying the single point of embedding.
Deoxyribonucleic acid (DNA) is utilized in the CBIR methodology that is proposed where the images are
initially stored in sequences of DNA after which the amino acid that is corresponding to it is extracted; this is
utilized as feature vectors. Here, the dimensionality reduction for feature vectors is achieved as well as the
required information is preserved [13]-[17]. Nayakwadi and Fatima [18], an image retrieval system that is
supervised weakly termed a class agnostic method is proposed based on CNN. The images in the database
have been pre-processed to split the background from the foreground and these foregrounds are stored as
clusters. Wang et al. [19], Gao et al. [20], and Zhu et al. [21], the focus of this paper is to decrease the
calculation on the count of data, which is performed on the online stage, and avoid any mismatches that occur
by mixing of backgrounds [22], [23].
In the last few years, the complexity of multimedia content, especially the images, has grown
exponentially, and on daily basis, more than millions of images are uploaded to different archives such as
Twitter, Facebook, and Instagram. Search for a relevant image from an archive is a challenging research
problem for the computer vision research community. Most search engines retrieve images based on
traditional text-based approaches that rely on captions and metadata. In the last two decades, extensive
3. Int J Reconfigurable & Embedded Syst ISSN: 2089-4864
Efficient content-based image retrieval using integrated dual deep convolutional (Feroza D. Mirajkar)
299
research is reported on CBIR, image classification, and analysis. In CBIR and image classification-based
models, high-level image visuals are represented in the form of feature vectors that consists of numerical
values. The research shows that there is a significant gap between image feature representation and human
visual understanding. This research work develops a dual-deep CNN for content-based image retrieval;
moreover, dual-deep CNN integrates two CNN. First CNN extracts the deep features, and this is given to the
second CNN that holds the characteristics of exploiting the semantic features concerning the dataset.
Furthermore, the novel directed graph is designed with two distinctive nodes known as memory node and
learning block node; in the case of a large dataset, a novel strategy is introduced for generating the efficient
feature. IDD-CNN is evaluated considering the two distinctive oxford and Paris dataset considering the
different metrics like mean average precision (mAP) and mean absolute error (mAE).
This research work is organized as shown in: section 1 starts with the background of image retrieval
and the importance of CBIR with deep learning. Further, a few existing models are discussed along with their
shortcomings. This section ends with research motivation and contribution. Section 2 discusses the proposed
architecture of IDD-CNN along with algorithm and mathematical modelling. Section 3 evaluates the IDD-
CNN based on the various difficulty level.
2. PROPOSED METHOD
Content-based image retrieval (CBIR) is a widely used technique for retrieving images from huge
and unlabeled image databases. Hence, the rapid access to these huge collections of images and the retrieving
of a similar image of a given image (Query) from this large collection of images presents major challenges
and requires efficient techniques. The performance of a content-based image retrieval system crucially
depends on the feature representation and similarity measurement. Figure 2 shows the proposed model
workflow; input image is given to the first deep-CNN to exploit the features and generated feature is given to
the second deep custom CNN that helps in exploiting more features. Moreover, a novel directed graph is
designed along with two distinctive blocks i.e. memory block and learning block.
Figure 1. Proposed workflow
Considering the M parameter in F dimensional node features denoted as Z belongs to Tm×F
; forward
propagation of the designed custom convolutional layer is computed as (1).
J(n+1)
= σ(F−1/2
CF−1/2
J(n)
Y(n)
) (1)
According to (1), J(n)
indicates the designed output layer of custom-CNN given as Z as the input which can
be given as J(q)
= Z. Furthermore, C indicates an adjacent matrix for the given graph-structured data. In
general, the adjacency matrix with the self-connection is defined through C = C + KP with Kp as the identity
matrix. F indicates the diagonal degree matrix along with its elementDi,j = ∑ Ck,l
l . Further, Y(n)
indicates the
weight matrix in FCN-layer; also σ indicates the activation function that indicates non-linearity in given
convolutional layers. Moreover, (1) can be further fragmented into the (2).
4. ISSN: 2089-4864
Int J Reconfigurable & Embedded Syst, Vol. 12, No. 2, July 2023: 297-304
300
J(n+1)
= σ(B(n)
Y(n)
) (2)
According to (2) forms the similar forward propagation as described in the earlier equation of FC-layer;
furthermore, nth layer output is recognized through a given weight matrix also known as the normalized
adjacency matrix R. J(n)
indicates the node feature vector that can be updated with the adjacency node feature
along with the matrix weight.
B(n)
= RB(n)
(3)
R indicates normalized adjacency matrix and given in (4).
R = F−1/2
CF−1/2
(4)
Moreover, considering the previous work on manifold mapping, this research paper develops an
integrated custom-CNN which tends to learn the novel feature representation and features are updated
through the corresponding neighboring node in a given database. Here Algorithm 1 shows the architecture of
IDD-CNN algorithm.
Algorithm 1. IDD-CNN algorithm
Input as memory block with its size o, learning block M, novel directed graph i and
training image set Kk
The expected output is learning block C and network W
Step1: Parameter Initialization
Step2: The exploitation of feature generation
Step3: Initiating the learning block Ed = [Ed
(0)
, Ed
(1)
,… … . . , Ed
(i−1)
through random process
mechanism
Step4: Designing of a memory block Eo = [Eo
(0)
,Eo
(1)
, … … . . , Eo
(i−1)
Step5: While iter is less than MIN do
Step6: Considering the mini-batch as the input along with CNN backbone as the output
Step7: Designing novel directed graph C= [C
(0)
, C
(1)
, … . . , C(i−1)
]
Step8: Considering each directed graph, other CNN outputs the updated feature
representations.
Step9: Network updation with backpropagation with designed objective
Step10: Learning block updation with algorithm considering the feature representation
Step11: Updation of memory block with optimized feature representation
Step12: End while loop
Step13: Return optimal network and learning blocks
Algorithm 1 shows the IDD-CNN algorithm where input is taken as the memory block-learning
block, designed directed graph along with training set. Further expected output includes the learning block
along with trained network. At first, the parameter is initialized, and features are exploited considering the
designed CNN. Further, we initiate the learning block and memory block is designed, later novel directed
graph is constructed and another CNN is utilized for the updated feature representation. Moreover, this
process is iterative approach hence other parameter are updated and learning blocks along with the trained
network are observed. Once the model is designed, it needs to be evaluated considering the benchmark
dataset. Evaluation is carried out in next section.
3. PERFORMANCE EVALUATION
Deep learning architecture like CNN has emerged as one of the major alternatives for hand-designed
features in the last decade; architecture like CNN automatically learns the feature from data. This research work
exploits and develops yet another CNN architecture to extract the optimal features. This section of the research
evaluates the IDD-CNN model considering the benchmark dataset of Oxford [24]. IDD-CNN is evaluated
considering the image retrieval and metrics evaluation. Also, comparison is carried out considering the ResNet
and VggNet model along with its various model including collaborative approach as discussed in [25].
3.1. Dataset details and system configuration
IDD-CNN is designed using python as a programming language along with various deep learning
libraries. System configuration includes 2 TB of the hard disk along with 4 GB CUDA enabled Nvidia
graphics. IDD-CNN is evaluated considering the standard dataset of oxford5k. Moreover, this research
utilizes the ROxford5K, which comprises 4,993 images along with 70 query images. Figure 3 shows the
5. Int J Reconfigurable & Embedded Syst ISSN: 2089-4864
Efficient content-based image retrieval using integrated dual deep convolutional (Feroza D. Mirajkar)
301
sample image of the Oxford dataset and Paris dataset; the first row shows the five samples of Oxford dataset
and the second row shows the five sample images of Oxford dataset.
Figure 3. Sample of Roxford and Rparis
3.2. Image retrieval and re-ranking
IDD-CNN performs the image retrieval based on the given query. This section evaluates the image
retrieval and reranking is carried out based on the relevance. Figure 4 shows the query image and Figure 5
shows the top 10 images ranked according to the relevance.
Figure 4. Query image
Figure 5. Top 10 images ranked according to the relevance
3.3. Metrics evaluation
The mAP or sometimes simply just referred to as AP is a popular metric used to measure the
performance of models doing document/information retrieval and object detection tasks.
6. ISSN: 2089-4864
Int J Reconfigurable & Embedded Syst, Vol. 12, No. 2, July 2023: 297-304
302
mAP = S(∑ average_precision(s)
S
s=1 )
−1
(5)
As shown in (5), S indicates the defined set for queries and average_precision(s) indicates the
average precision for given query s. mAP metrics is one of the important metrics where for given query
average precision is computed; later mean of all these average precision gives the single number known as
mAP that shows the performance of the model at the given query.
3.4. Comparison with ResNet based architecture
Figure 6 shows the comparison of various ResNet architecture with proposed architecture IDD-CNN
considering the difficulty level medium. In here, R-Gem achieves 64.7, its query expansion version R-Gem+
QE achieves 67.2. Further R-Gem along with deep spatial matching (DSM) approach R_Gem+DSM achieves
65.3. Similarly, R-Gem+DFS achieves mAP of 69.8; furthermore, existing model collaborative approach
observes 66. However, in comparison IDD-CNN achieves 74.37. Figure 7 shows the comparison of various
ResNet architecture with proposed architecture IDD-CNN considering the difficulty level hard. Figure 2
shows the comparison of various ResNet architecture with proposed architecture IDD-CNN considering the
difficulty level medium. Here, R-Gem achieves 38.5, and its query expansion version R-Gem+QE achieves
40.8. Further R-Gem along with the DSM approach R_Gem+DSM achieves 39.2. Similarly, R-Gem+DFS
achieves mAP of 40.2; furthermore, the existing model collaborative approach observes 42.5. However, in
comparison IDD-CNN achieves 57.7.
Figure 6. ResNet architecture comparison with the
proposed model on medium difficulty level
Figure 7. ResNet architecture comparison with the
proposed model on hard difficulty level
3.5. Comparison with VGGNet based architecture
Figure 8 shows the comparison of various ResNet architecture with proposed architecture IDD-CNN
considering the difficulty level medium. Further, Figure 4 shows the comparison of various VGGNET
architectures with the proposed architecture IDD-CNN considering the difficulty level medium. Here, V-
Gem achieves 38.5, and its query expansion version R-Gem+QE achieves 40.8. Further R-Gem along with
the DSM approach R_Gem+DSM achieves 39.2. Similarly, R-Gem+DFS achieves mAP of 40.2;
furthermore, the existing model collaborative approach observes 42.5. However, in comparison IDD-CNN
achieves 57.
Figure 9 shows the comparison of various ResNet architecture with proposed architecture IDD-CNN
considering the difficulty level hard comparison of various VGGNET architecture with proposed architecture
IDD-CNN considering the difficulty level medium. Here, V-Gem achieves 38.5, and its query expansion
version R-Gem+QE achieves 40.8. Further R-Gem along with the DSM approach R_Gem+DSM achieves
39.2. Similarly, R-Gem+DFS achieves mAP of 40.2; furthermore, the existing model collaborative approach
observes 42.5. However, in comparison IDD-CNN achieves 57.7.
3.6. Comparative analysis
In the earlier section, Figure 2 and Figure 3 shows the comparison on various ResNet architecture
including the existing collaborative model. Moreover, comparative analysis suggests that IDD-CNN
improvises the model with 6.54% from R-Gem+DFS. Considering medium level and considering difficulty
level as hard, IDD-CNN achieves marginal improvisation of 35.76%. Similarly, considering the VGGNet
model, IDD-CNN achieves 4.59% of improvisation in comparison with V-Gem+DFS+Collaborative.
Moreover, considering the difficulty level as hard, IDD-CNN achieves 24.08% improvisation than the
existing model.
7. Int J Reconfigurable & Embedded Syst ISSN: 2089-4864
Efficient content-based image retrieval using integrated dual deep convolutional (Feroza D. Mirajkar)
303
Figure 8. VggNet architecture comparison with the
proposed model on medium difficulty level
Figure 9. VggNet architecture comparison with the
proposed model on hard difficulty level
4. CONCLUSION
CBIR is one of the critical task and it has emerged yet another popular task in image processing due
to enormous growth in multimedia like image. This research work designs and develops an efficient retrieval
and ranking mechanism named integrated dual deep convolutional neural network (IDD-CNN). IDD-CNN is
evaluated on image retrieval and metrics evaluation; at first is image retrieval and re-ranking based on
defined query. Later, IDD-CNN is evaluated considering mAP metrics; further evaluation is carried out by
comparing IDD-CNN with various CNN model and its variant architecture. IDD-CNN is proven efficient and
highly improvised, thus future scope lies in reducing the image retrieval time and varying the different
dataset.
REFERENCES
[1] L. Song, Y. Miao, J. Weng, K. K. R. Choo, X. Liu, and R. H. Deng, “Privacy-preserving threshold-based image retrieval in cloud-
assisted internet of things,” IEEE Internet of Things Journal, vol. 9, no. 15, pp. 13598–13611, Aug. 2022, doi:
10.1109/JIOT.2022.3142933.
[2] Z. Xia, L. Jiang, D. Liu, L. Lu, and B. Jeon, “BOEW: a content-based image retrieval scheme using bag-of-encrypted-words in
cloud computing,” IEEE Transactions on Services Computing, vol. 15, no. 1, pp. 202–214, Jan. 2022, doi:
10.1109/TSC.2019.2927215.
[3] K. N. Sukhia, S. S. Ali, M. M. Riaz, A. Ghafoor, and B. Amin, “Content-based image retrieval using angles across scales,” IEEE
Geoscience and Remote Sensing Letters, vol. 19, pp. 1–5, 2022, doi: 10.1109/LGRS.2021.3131340.
[4] S. R. Dubey, “A decade survey of content based image retrieval using deep learning,” IEEE Transactions on Circuits and Systems
for Video Technology, vol. 32, no. 5, pp. 2687–2704, May 2022, doi: 10.1109/TCSVT.2021.3080920.
[5] A. P. Byju, B. Demir, and L. Bruzzone, “A progressive content-based image retrieval in JPEG 2000 compressed remote sensing
archives,” IEEE Transactions on Geoscience and Remote Sensing, vol. 58, no. 8, pp. 5739–5751, Aug. 2020, doi:
10.1109/TGRS.2020.2969374.
[6] O. E. Dai, B. Demir, B. Sankur, and L. Bruzzone, “A novel system for content-based retrieval of single and multi-label high-
dimensional remote sensing images,” IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, vol.
11, no. 7, pp. 2473–2490, Jul. 2018, doi: 10.1109/JSTARS.2018.2832985.
[7] J. Xu, C. Wang, C. Qi, C. Shi, and B. Xiao, “Iterative manifold embedding layer learned by incomplete data for large-scale image
retrieval,” IEEE Transactions on Multimedia, vol. 21, no. 6, pp. 1551–1562, Jul. 2019, doi: 10.1109/TMM.2018.2883860.
[8] A. Raza, H. Dawood, H. Dawood, S. Shabbir, R. Mehboob, and A. Banjar, “Correlated primary visual texton histogram features
for content base image retrieval,” IEEE Access, vol. 6, pp. 46595–46616, 2018, doi: 10.1109/ACCESS.2018.2866091.
[9] U. Chaudhuri, B. Banerjee, and A. Bhattacharya, “Siamese graph convolutional network for content based remote sensing image
retrieval,” Computer Vision and Image Understanding, vol. 184, pp. 22–30, Jul. 2019, doi: 10.1016/j.cviu.2019.04.004.
[10] Y. Li, X. Kong, L. Zheng, and Q. Tian, “Exploiting hierarchical activations of neural network for image retrieval,” in MM 2016 -
Proceedings of the 2016 ACM Multimedia Conference, Oct. 2016, pp. 132–136, doi: 10.1145/2964284.2967197.
[11] Y. Gong, L. Wang, R. Guo, and S. Lazebnik, “Multi-scale orderless pooling of deep convolutional activation features,” in Lecture
Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol.
8695 LNCS, no. PART 7, 2014, pp. 392–407, doi: 10.1007/978-3-319-10584-0_26.
[12] Z. Zhang, L. Wang, Y. Wang, L. Zhou, J. Zhang, and F. Chen, “Dataset-driven unsupervised object discovery for region-based
instance image retrieval,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 45, no. 1, pp. 247–263, Jan.
2023, doi: 10.1109/TPAMI.2022.3141433.
[13] Y. Zeng et al., “Keyword-based diverse image retrieval with variational multiple instance graph,” IEEE Transactions on Neural
Networks and Learning Systems, pp. 1–10, 2022, doi: 10.1109/TNNLS.2022.3168431.
[14] J. Pradhan, C. Bhaya, A. K. Pal, and A. Dhuriya, “Content-based image retrieval using DNA transcription and translation,” IEEE
Transactions on NanoBioscience, vol. 22, no. 1, pp. 128–142, Jan. 2023, doi: 10.1109/TNB.2022.3169701.
[15] S. Yang, L. Ma, and X. Tan, “Weakly supervised class-agnostic image similarity search based on convolutional neural network,”
IEEE Transactions on Emerging Topics in Computing, vol. 10, no. 4, pp. 1789–1798, Oct. 2022, doi:
10.1109/TETC.2022.3157851.
[16] F. Mirajkar and R. Fatima, “Content based image retrieval using the domain knowledge acquisition,” in Proceedings of the 3rd
International Conference on Inventive Research in Computing Applications, ICIRCA 2021, Sep. 2021, pp. 509–515, doi:
10.1109/ICIRCA51532.2021.9544821.
8. ISSN: 2089-4864
Int J Reconfigurable & Embedded Syst, Vol. 12, No. 2, July 2023: 297-304
304
[17] L. Math and R. Fatima, “Adaptive machine learning classification for diabetic retinopathy,” Multimedia Tools and Applications,
vol. 80, no. 4, pp. 5173–5186, Feb. 2021, doi: 10.1007/s11042-020-09793-7.
[18] N. Nayakwadi and R. Fatima, “Automatic handover execution technique using machine learning algorithm for heterogeneous
wireless networks,” International Journal of Information Technology (Singapore), vol. 13, no. 4, pp. 1431–1439, Aug. 2021, doi:
10.1007/s41870-021-00627-9.
[19] L. Wang, X. Qian, Y. Zhang, J. Shen, and X. Cao, “Enhancing sketch-based image retrieval by CNN semantic re-ranking,” IEEE
Transactions on Cybernetics, vol. 50, no. 7, pp. 3330–3342, Jul. 2020, doi: 10.1109/TCYB.2019.2894498.
[20] J. Gao, X. Shen, P. Fu, Z. Ji, and T. Wang, “Multiview graph convolutional hashing for multisource remote sensing image
retrieval,” IEEE Geoscience and Remote Sensing Letters, vol. 19, pp. 1–5, 2022, doi: 10.1109/LGRS.2021.3093884.
[21] L. Zhu, H. Cui, Z. Cheng, J. Li, and Z. Zhang, “Dual-level semantic transfer deep hashing for efficient social image retrieval,”
IEEE Transactions on Circuits and Systems for Video Technology, vol. 31, no. 4, pp. 1478–1489, Apr. 2021, doi:
10.1109/TCSVT.2020.3001583.
[22] W. Xiong, Z. Xiong, Y. Zhang, Y. Cui, and X. Gu, “A deep cross-modality hashing network for SAR and optical remote sensing
images retrieval,” IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, vol. 13, pp. 5284–5296,
2020, doi: 10.1109/JSTARS.2020.3021390.
[23] M. Zhang, Q. Cheng, F. Luo, and L. Ye, “A triplet nonlocal neural network with dual-anchor triplet loss for high-resolution
remote sensing image retrieval,” IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, vol. 14, pp.
2711–2723, 2021, doi: 10.1109/JSTARS.2021.3058691.
[24] F. Radenovic, A. Iscen, G. Tolias, Y. Avrithis, and O. Chum, “Revisiting Oxford and Paris: large-scale image retrieval
benchmarking,” in 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition, Jun. 2018, pp. 5706–5715, doi:
10.1109/CVPR.2018.00598.
[25] J. Ouyang, W. Zhou, M. Wang, Q. Tian, and H. Li, “Collaborative image relevance learning for visual re-ranking,” IEEE
Transactions on Multimedia, vol. 23, pp. 3646–3656, 2021, doi: 10.1109/TMM.2020.3029886.
BIOGRAPHIES OF AUTHORS
Prof. Feroza M. Mirajkar is working as Assistant Professor in Department of
E&CE, in Khaja Banda Nawaz College of Engineering, Kalaburagi, Karnataka, India since
2013. Completed Master of Technology from Basaveshwar College of Engineering, Bagalkot,
Karnataka, India. She has attended many workshops such as “Innovative research techniques”
a National workshop, CUK, Kalaburagi, participitated in a Two-week ISTE STTP on
“Pedagogy for Effective use of ICT in Engineering Education” conducted by Indian institute of
Institute of Technology Bombay at NK Orchid college of Engineering and Technology,
Solapur and many more. She has 12 publications which include both national and international
journal and conferences with one IEEE. She can be contacted at email: mmferoza@gmail.com.
Dr. Ruksar Fatima is presently working as Dean Faculty of Engineering and
Technology, Khaja Bandanawaz University, Gulbarga with an experience of 18 years as a
dedicated, resourceful education professional. She has received 4 awards, Award for Best
Scientific publication by VGST in 2018, RURLA award for Distinguished Scientist in 2018,
Chairman for IETE Gulbarga sub-center and Best Senior Researcher (Female) by international
Academic and Research Excellence Awards 2019. She has 31 international publications in
reputed journals and technical member and reviewer of many famous international journals.
She can be contacted at email: ruksarf@gmail.com.
Dr. Shaik A. Qadeer is currently working as Professor in the Department of
Electrical and Electronics Engineering at MJ College of Engineering, Hyderabad, Telangana
since May 2015. He is the active member of the various Professional Societies and awardee for
academic excellence during his Bachelor and master’s degree studies including 5th
rank holder
in the University. He is having 18 Scopus indexed publications, 8 Web of Science publications
and one Indian patent. The author having 22 years of teaching experience and his interest
include industrial automation, cyber physical systems, signal processing and machine learning.
He is a member of IEEE. He can be contacted at email: haqbei@gmail.com.