Search

Alzheimer's disease (AD) is a type of neuron disease; its nature causes the brain cells to degenerate and die. It's a progressive disorder. It's an incurable disease and develops memory impairment as it progresses. The precise diagnosis of AD plays a vital role in the patient's health care, especially at its initial stage. The early detection of the disorder can help the patient get proper treatment and prevent further irreversible damage to the brain. This paper focuses on a comprehensive study of computer-aided diagnosis of AD by Convolution Neural Network (CNN). Many researchers have performed various ways through CNN for predicting AD at its dawn stage. The comprehensive study of paper shows various algorithms in CNN for early diagnosis of the disease through neuroimaging biomarkers [ABSTRACT FROM AUTHOR]

In the recent times, deep learning has gained significant importance in various domains. Computer Vision is one the domains in which deep learning has seen vast development and is still gaining a lot of attention. CNNs are deep architectures that are specifically applied for image classification and object detection in computer vision. Image classification has also been successfully applied in the field of agriculture. Using deep learning, various agricultural problems have been addressed and solved. Keeping in view the enormous uses of deep learning in agricultural domain, this paper gives a brief review of the various works proposed by the researchers working in the field of deep learning and agriculture. This paper will put some light on the various agricultural problems, the deep learning models that have been employed to solve the problem, any pre-processing techniques that have been applied on the images to prepare the dataset for the specific model and the results that have been achieved by the model in solving the problem they focus on. It will also provide any research gap found in the specific work. This work will be useful for future researchers to gain an overview of the work done so far and the deep learning techniques used in agriculture. [ABSTRACT FROM AUTHOR]

This paper mostly focuses on real-time detection and classification of biomedical waste objects. Traditional methods for detecting the objects like biomedical waste objects in machine learning are substituted by recent technology of object detection methods in deep learning by building Convolutional Neural Network (CNN) which is nothing but the components of deep learning. This paper proposes a deep learning approach for object detection of biomedical waste material by evaluating the TensorFlow object detection API, which can used for real time application, this biomedical object detection API further we can used for automate the biomedical waste segregation system, this proposed system has been trained using more than 500 images for each type of biomedical waste object. Tensorflow is useful and easy to object detection that can form powerful image recognition software, in this proposed work we are using Faster R-CNN algorithm for developing this object detection process. [ABSTRACT FROM AUTHOR]

Now a day there is exponential growth in music files, so the need for automatic music genre classification approach is also increasing. Current classification systems of music genres developed using machine learning algorithms and some of the limitations of existing music genre systems are doesn't work with large database, low accuracy etc. In this paper, presented a large music dataset namely GTZAN with 10 distinct music genres like jazz, rock, hip-hop etc. In order train the classification model of music genre system, a deep learning based technique called CNN (Convolution Neural network) is applied. The proposed method performs the classification of music genre into innumerable genre with the help of feature vector extraction, which plays a vital role in training the model. The feature vector called MFCC (Mel Frequency Cepstral Coefficient) considered in classifying the raw files which are converted to respective music genres or types. The performance of various classification techniques are analyzed compared with respect to evaluation metrics. [ABSTRACT FROM AUTHOR]

Now-a-days, Breast Cancer is a serious cause in world and it raises a death ratio in drastic manner. Most of the women are suffered due to this kind of breast cancer disease, in which several literatures are available regarding this breast cancer identification, but all are strucked over certain level. Lots of researchers in industry providing different digital image processing schemes to identify the Breast Cancer in clear manner to save lives of women but all are strucked with real world prediction scenarios. For Breast Cancer identification mammography is a well-known technique, in which it screens the breast cancer cells in clear manner. However, based on the workload and manual expertise failures, a numerous failures and wrong classification procedures reported as well. So, that a novel computer assisted digital image processing technique and smart tool is required to provide intelligent prediction logics to identify the breast cancer on earlier conditions instead of detecting that in complex stages. In this paper, a novel Deep Learning enabled Image Processing tool is designed, called as Digital Image Processing Scheme enabled SmartKit (DIPSK) to identify the breast cancer disease effectively on earlier stages without any human intervention. The prop osed DIPSB enables the system to acquire the input scanning image and process that with DIP techniques with respect to pre-processing feature extraction and classification. The detailed features are cross-validated with the trained samples, in which the Deep Learning procedure called Improved Convolutional Neural Network (iCNN) is utilized by the SmartKitto provide classification process in accurate manner. The final resulting of this SmartKitis more efficient as compare to the traditional image processing schemes and provides accurate prediction on earlier stages. The resulting section provides the proper proof for the classification accuracy and prediction logics. [ABSTRACT FROM AUTHOR]

Stress is the regular response of the body in demanding situations but the chronic stress is very harmful affecting our emotional and physical well-being. The role of Autonomic Nervous System (ANS) in stress is discussed. In this survey various paper based on the stress detection using Convolutional Neural Networks (CNN) that were published in the range of 4 years (2019 – 2022) were collected. The stress detection processes are classified based on different sources of information used. The sources of data include physiological signals, report collected through questionnaire sessions, data from observing the actions in social media and more. Among various bio-signals, stress detection from Electrocardiograph signal is mainly focused and discussed in this survey. This paper analyses each article focusing on the four major factors namely the stressor used, pre-processing of the datasets collected, methodologies proposed, the outcome of the various papers. [ABSTRACT FROM AUTHOR]

State-of-the-art solutions in the areas of language modelling, generating text, speech recognition, generating image descriptions or video tagging, have been using Recurrent Neural Networks as the foundation for their approaches. Understanding the underlying concepts is therefore of tremendous importance if we want to investigate frontier methods and develop new approaches for building smart digital solutions. This paper proposes a smart city control system architecture based on deep convolutional neural networks. The control system has a multilayer architecture that combines loosely coupled intelligent components. As the main layer, a solution based on deep learning technology is applied, which allows solving several tasks simultaneously: segmentation, detection and classification of images received from surveillance cameras of the smart city system. The data obtained at the output of this layer is used for further analysis and decision-making in the smart city system. The proposed architecture has a high degree of modularity and allows the replacement of individual elements in a loosely coupled architecture. In this paper, deep learning and computer vision technologies are also considered, on the basis of which the image processing layer from video cameras is implemented. A masked recurrent neural network is used for this task. [ABSTRACT FROM AUTHOR]

The objective of this paper is to identify the signature verification system which can overcome individual presence for verification and recognition task of handwritten signature. Amid corona duration it has been found that there is a need of automatic online system which can omit physical presence of person for signature verification. This paper has proposed signature verification model with better accuracy. Novel Architecture of convolution neural network is proposed which exhibited 99% training accuracy on GPDS synthetic Signature database. Writer independent Features of signature were extracted through CNN and these features were fed into machine learning model for classification. Novel architecture is proposed in this paper. [ABSTRACT FROM AUTHOR]

Effective traffic prediction is crucial due to a surge in deliveries by commerce and urbanization. This has led to a notable rise in traffic within megacities, causing route delays to the final destinations and countless vehicle accidents. E-commerce has been in a constant boom, as buying something online and having it delivered to the front door is easier than going to the store. As more people engage in this activity, e-commerce platforms' challenges are more complicated and need to be addressed faster. However, these challenges escape the delivery company's scope when external factors influence the objective of optimized deliveries, for example, traffic issues or bad weather during the last mile, issues that are only exacerbated where traffic sensors are not widely used (i.e., underdeveloped countries). The main contributions of this research are to (1) provide a contextual foundation of current frameworks used for traffic prediction, (2) use social media and multi-modal traffic-related data (weather, points of interest, calendar of events) by leveraging social network analysis to improve the accuracy of traffic prediction, and (3) to show a methodology that can be used for partially observed traffic. The proposed methodology includes deep learning tools like Long-Short Term Memory Networks, attention mechanisms, Graph Convolutional Networks, and social media tools like sentiment analysis. [ABSTRACT FROM AUTHOR]

In computer vision, object detection is an essential task to identify and classify objects in the image or video. The recent advancements in deep learning and convolutional Neural Networks (CNNs) have significantly improved the performance of object detection techniques. In an unconstrained environment, the study in this paper provides a detailed analysis of object detection techniques and various challenges, datasets, or state-of-the-art approaches. In addition, a comparative analysis of these methods is presented and its strengths and weaknesses are highlighted. Finally, we've provided some new research directions for improving the detection of objects in uncontrolled environments. [ABSTRACT FROM AUTHOR]

Diseases can have a detrimental effect on plants depreciating their market value. Disease diagnosis is important for improving plant yield. Over the years, farmers are relying on Soft Computing (SC) techniques for the purpose of early diagnosis of various plant diseases. The paper builds an application using a Deep Learning (DL) model for timely disease diagnosis. The research employs pre-trained Convolution Neural Networks (CNN) for identifying crop diseases and insect pests. CNN efficiently tackles the challenges related to image stability and the number of parameters. PlantVillage dataset is used for training and testing the model. The dataset comes with fifteen different classes of infected and healthy leaves. The work applies image pre-processing techniques and data augmentation for enhancing the image recognition rate. The paper implements a crop disease recognition model comprising of Inception-ResNet-v2, Visual Geometry Group 16, DenseNet201 and, Inception_V3 for training. The model and subsequently the application are validated and tested using standard methods. The convolution layer of the networks uses a non-linear activation function-ReLU which has gained popularity in the deep learning domain for its simple activation strategy. Later the name of the crop reporting the highest matching degree is displayed. Lastly, the accuracy of all the networks is compared against the traditional convolution network. It is inferred that the system thus aims to efficiently identify crop diseases and produces better recognition accuracy than the traditional convolution networks. [ABSTRACT FROM AUTHOR]

Trash classification is an important activity which helps in identification of waste. In this paper, a Classical- Quantum Transfer learning model, namely TrashQNet is proposed to classify trash into two classes, Organic and Recyclable trash. Classical-Quantum Transfer learning is a combination of machine learning and quantum computing. The proposed TrashQNet model uses a pre-trained DenseNet169 network for the feature extraction process and a variational quantum circuit as a classifier. The performance of TrashQNet is compared with classical machine learning models K- Nearest Neighbour & Support Vector Machine, deep learning model - Convolutional Neural Network and transfer learning model. TrashQNet outperforms all these models, it achieves an accuracy of 94% on test dataset. [ABSTRACT FROM AUTHOR]

With an increase in the online picture database, the need for gender and age classification has become an active area of research. It helps in detection and investigation purposes. With the emergence of computer vision, deep learning (DL) is being applied in research due to its wide range of real-world applications in image classification. This paper investigates the performance of two DL models, the Convolution Neural Network (CNN) and the Residual Neural Network (ResNet18) model, for analyzing their efficacy in predicting gender and age classification tasks using facial images. The Adience dataset, a widely used benchmark dataset for these tasks, is used to train and test the respective models. This study uses a 3-fold empirical model evaluation on CNN and ResNet18 model, and results were found on three main assessment criteria-age accuracy, 1-off accuracy, and gender detecting accuracy. The ResNet18 model slightly outperformed the CNN model in both functions by showing a 94% exact match on age classification and 92% on gender prediction. This paper highlights the effectiveness of DL models in gender and age classification tasks using facial images and the importance of the model architecture and design in achieving high accuracy and robustness. [ABSTRACT FROM AUTHOR]

People in Tamil Nadu have been more interested in history during the past years. When people see the old script inscribed on stone walls, potteries, copper plates, and Palm Manuscripts, history is more intriguing. But people have to learn old scripts in order to read these inscriptions, so not many people are capable of reading them. A ground breaking effort that educates the public about these ancient scripts and eventually sparks their interest in history is the use of deep learning algorithms for script recognition. Recognizing ancient Tamil characters is a difficult challenge for epigraphers because the language has evolved over the previous three thousand years with distinct characters set. Though there have been a few studies published on the recognition of ancient Tamil alphabets, there have been no papers on the recognition of ancient Tamil numbers. This project's main objective is to use convolution neural network (CNN) to recognize Tamil digits used between 9th to 12th century in Tamil inscriptions. In this paper work, the inscription photographs acquired from various Tamil temples are pre-processed. The color photographs were cropped, converted to grayscale and then to binary images during the pre-processing process. These pre-processed images are used as training and validation datasets for the model. [ABSTRACT FROM AUTHOR]

Automated vehicle number plate recognition or Number Plate Recognition (NPR) is crucial in various applications, including law enforcement, traffic management, and toll collection systems. However, the traditional method of manual inspection of images is time-consuming and prone to errors. With the advancements in computer vision and deep learning techniques, identifying vehicle number plates has become more efficient and accurate. In this paper, we present a methodology for identifying vehicle number plates using the VGG16 convolutional neural network. We collected a dataset of vehicle images with their corresponding number plates and preprocessed the data to train the CNN using algorithms like transfer learning. The VGG16 CNN has been pre-trained on a large data set that is the ImageNet dataset, which makes it an excellent candidate for fine-tuning on specific tasks like vehicle number plate recognition. After training, we evaluated the performance of the CNN on a test dataset and deployed it in a real-world application. Our results show that the VGG16 CNN can accurately recognize number plates on different types of vehicles in various lighting conditions. This methodology improves efficiency and accuracy while reducing manual inspection errors. It can be used as an essential tool for automated number plate recognition in various applications, enabling faster processing times and reducing costs associated with manual inspection. The methodology presented in this paper demonstrates the effectiveness of deep learning techniques in automating vehicle number plate recognition. The VGG16 CNN architecture, combined with transfer learning and preprocessing techniques, can accurately recognize number plates, providing a reliable and efficient solution for automated number plate recognition system. [ABSTRACT FROM AUTHOR]

Water quality prediction has a crucial significance in ensuring the safety and usability of water resources. In this paper, we propose a CNN with BiLSTM model for accurate water quality prediction. The model combines convolutional neural network layers with BiLSTM to effectively capture the spatio-temporal dependencies present in the data. Experimental results on the water quality dataset demonstrate the superior performance of the proposed model. The model achieves a validation accuracy of 92.83%. These results surpass those obtained by other models such as GRU, LSTM, and CNN with LSTM. The CNN with BiLSTM model's ability to effectively learn and classify water samples based on provided features highlights its potential for reliable water quality prediction. This paper contributes to the field of water quality prediction by providing a robust model that can aid in ensuring the safety and suitability of water resources. [ABSTRACT FROM AUTHOR]

In today's world which is digitizing increasingly, facial expressions serve as one of the vital indicators of variety of human emotions. With the advancement of technologies like Artificial Intelligence (AI) systems, even machines can process and understand these facial expressions with wide number of applications across various domains. However, the challenge still lies in accounting the human diversity, which leads to distinct characteristics for each individual. And an attempt of creating a single model to fit all such individuals becomes nearly impractical. To tackle this challenge, an organized approach based on inter country facial distinctions could be a feasible solution. This approach works better within the specific country, but the universality is also required which is again a very big challenge. For this as a solution a collaborative methodology is being proposed. The aim of this paper is to prove that it is possible to create a universal model for detecting facial expressions with higher precision and accuracy by combining various datasets from different nationalities to obtain a single dataset with uniform distribution. This paper does so by using datasets of two countries i.e., JAFFE and CK datasets and generating a convolutional neural network-based model by optimizing the trade-offs between model complexity, training time, and required computational resources. [ABSTRACT FROM AUTHOR]

The use of a bioinspired Convolutional Neural Network (CNN) model in this paper's unique automated method for malaria diagnosis in blood smear images. Extensive development in computer-aided diagnosis systems has been driven by the urgent need to fight malaria, a disease that poses a serious threat to life. However, the precision, accuracy, and recall rates of current approaches are constrained, which reduces their usefulness in real-world settings. The suggested approach places a strong emphasis on incorporating bioinspired ideas into the CNN model design to overcome these limitations. The bioinspired CNN model outperforms conventional CNN architectures in terms of greater specificity and expanded capabilities by taking inspiration from nature-inspired algorithms. This concept provides a considerable improvement over current methodologies by enabling more precise and reliable malaria detection process. The strategy that is being discussed combines machine learning strategies with extensive image processing capabilities. Through the use of several approaches, this integration makes it possible to provide a comprehensive solution to the problem of malaria detection. In comparison to conventional methods, experimental results show the proposed method to be effective, with precision rates of 3.9%, accuracy rates of 2.9%, and a surprising 3.5% increase in recall levels. Thus, this paper demonstrates the need for a novel strategy to malaria detection in light of the shortcomings of current techniques. By combining a bio-inspired CNN model with image processing and machine learning methods, it is possible to greatly improve the precision, accuracy, and recall rates for automated malaria identification in blood smear images. The results of this work show significant promise for developing the discipline of medical image analysis and supporting continuing global efforts to eradicate malaria for real-time scenarios. [ABSTRACT FROM AUTHOR]

Early identification and accurate diagnosis are crucial for successful therapy and improved patient outcomes of brain tumors. The study looks at how deep learning may help detect brain tumors, including how convolutional neural networks (CNNs) can change the face of diagnostic imaging. This underscores the need of precise and timely identification of brain cancer for efficient therapy and enhanced patient results. The study discusses the challenges associated with tumor segmentation and classification, such as variations in size, shape, and location. To overcome these challenges, deep learning techniques are utilized to analyze brain scans and identify tumor abnormalities. However, the paper underscores the importance of these models serving as tools to assist medical professionals rather than replacing their expertise. Additionally, the paper presents a comparative study that evaluates different techniques, artificial neural networks (ANN) achieve the highest accuracy at 99%, followed by Support Vector Machine (SVM) at 98.9%. vision transformer (ViT), Extreme Gradient Boosting (XG Boost), CNN-based dense Efficient Net, and Deep Neural Network (DNN) also show high accuracies above 98%. However, U-Net, Alex Net, hybrid ensemble method (Random Forest (RF), K-Nearest Neighbour (KNN), Decision Tree (DT), and Region-based convolutional neural networks (R-CNN)) have lower accuracies ranging from 91.66% to 97.305%. Overall, the findings highlight the potential of deep learning in enhancing the efficiency, precision of brain tumor diagnosis, leads to earlier treatment and enhanced patient prognosis. [ABSTRACT FROM AUTHOR]

This research proposes a unique approach to human action recognition using mobile sensor data and a computationally efficient 1D Convolutional Neural Network (1D-CNN). In this research, a 1D-CNN model is constructed to recognize human actions using data from an accelerometer sensor. In order to increase model performance, the study looks into the ideal number of layers and epochs. The proposed method also automates the data annotation to simplify the training process. This research highlights the significance of the chosen model and the size of segments in action recognition. This paper investigates the ideal segmentation size for precisely identifying actions. Experimental analysis confirms the effectiveness of the segmentation length in recognizing human actions by reducing false alarms. This paper demonstrates that increasing the number of fully connected layers does not increase precision or accuracy. The paper concludes by proposing subject-independent methods for action recognition and optimizing power consumption for wearable devices. It highlights the potential of using mobile sensor data and 1D-CNNs for future research in human action recognition. The method and model presented in this paper achieve 95% of recognition accuracy. [ABSTRACT FROM AUTHOR]

The enormous benefits and applications of Image classification and recognition are umpteen. Machine learning algorithms and Deep Neural Networks are like windfall to fathom the objective proficiently in streamlined manner. The prevalent improvement in this technology is that these networks do not call for any prior blueprints in terms of algorithms as prerequisites. The presented paper is an attempt to create a Convolutional Neural Network from scratch to classify the images from the well-known dataset – Cats and Dogs into their relevant baskets. Manifold open source accessible approaches to amplify the efficiency of the network are no more onerous. Further, data augmentation technique boosts the efficiency tremendously by extending the dataset with reoriented features from the same images. To untangle the same problem, Transfer Learning is also a compelling technique in which all the layers, neurons in each layer, weights of each neuron and all other parameters are predefined and we can amend the output layer as per the classes in the respective problem statement. In the present paper, we have tried to obtain a comparable efficiency with a significant reduction in parameters. [ABSTRACT FROM AUTHOR]

The importance of medicinal plants has widely increased throughout the world, especially during the COVID-19 pandemic. It is essential for a layman to identify a medicinal plant and know about its medicinal values. In this view, this paper proposes an approach that helps a person to identify the medicinal values of different plants. The paper focuses on designing a mobile application that can identify the leaf and list its medicinal values with a mere snapshot of it by performing instance segmentation, feature extraction and classification in the backend. The models were trained and tested on a custom dataset generated by collecting around 1500 images spreading over 5 medicinal plant species. The paper uses a deep learning-based segmentation approach for real-time segmentation of the leaf from the complex natural background. The segmented leaf will undergo a feature extraction process and will be given as an input to the traditional Machine Learning algorithms. Manual Feature Extraction and Deep Learning based Feature extraction processes were implemented. Various features including shape, colour and texture were extracted in the manual feature extraction process and a MobileNet pre-trained model was used for deep learning-based feature extraction. The Mask Region-based Convolutional Neural Network (Mask R-CNN) yielded the best performance for real-time segmentation with a Dice score of 0.92 and for the classification MobileNet + Logistic Regression gave the best accuracy of 0.97 and AUC score of 0.99. [ABSTRACT FROM AUTHOR]

Detection of disease in plants is a paramount issue in the agricultural sector of any country. Crop failure adversely affects the farmer and has an enormous impact on the economy. Modern technology has developed various tools using deep learning techniques to perform errorless predictions and to originate a rapid treatment. The purpose of this paper is to propose a deep learning based method to detect disease in plants using leaf images. Our main goal is to develop an efficient methodology to provide reliable predictions. We have considered CNN (Convolution Neural Network) algorithm as our principal detector. The proposed system can approach complex scenarios from the plant's body. The result achieved is very promising of 98.33 accuracy and 0.0047 loss on validation data. [ABSTRACT FROM AUTHOR]

Person detection is an interesting machine learning application that can be implemented in embedded devices. One such need is to automate an alarm system to detect whether a person that works in a Point of Sale is indeed working at his/her working time. This presence monitoring system can be automated by using an esp-eye development board by Espressif. In this paper, the key important aspect of modelling the person detection behavior is discussed and implemented in ESP32 technologies. The rate of accuracy in detecting a person's presence is up to 65% based on a four (4) minute window-time, and therefore it is reliable to be used as a person's monitoring automated system, based on the current requirement from the food industry. The inference time is circa 0.7 MS, that is the time between each sample in a situation whether a person is in front of the camera or not. [ABSTRACT FROM AUTHOR]

At present, detection and recognition of animals remain a complicated task and no distinctive technique can solve all situations strongly and effectively. Animal detection in images is challenging because of the complexity of wild environments. Including several different things: The large differences in shapes and color appearances of variations of objects which belong to the same class, and Lighting/Illumination Conditions. This paper presents an effective and efficient animal detection system for Birds detection in images. MobileNet SSD model is a Single Shot Multibox Detection (SSD) suggested in this work, which is capable of handling different shapes and view angles of the objects. This approach allows extracting features of image regions and using deep learning techniques for classifying regions into two categories: (bird and non-animal). The experimental results demonstrated the efficiency of classification and detection of single and multi-animals even in difficult environmental conditions. The accuracy of detection and classification is up to 98.8%. This methodology is more accurate than other similar approaches. This paper exhibits a method for training a convolutional neural network (CNN) based on object detection classifiers. [ABSTRACT FROM AUTHOR]

The rise of cyber-attacks such as Distributed Denial of Service (DDoS) and SQL injection has become a significant concern for organizations and individuals who rely on the internet. Traditional detection methods have become increasingly ineffective in addressing these challenges, necessitating the development of new and innovative solutions. This paper proposes using Convolutional Neural Networks (CNNs) to detect DDoS and SQL injection attacks. Our paper proposes a Convolutional Neural Network model with a multi-layer neural and relu activation function optimizer that reaches a higher degree of precision than previous Deep Learning models. In this study, we tested the relatively new dataset CIC-IDS-2018, which contains different types of attacks. With this dataset, our model achieves an unprecedented accuracy of>96%, minimizing computational time. [ABSTRACT FROM AUTHOR]

The detection of chest diseases from X-ray images is a crucial aspect of medical diagnostics, playing a vital role in the early identification and treatment of respiratory conditions. This review paper gives a general overview of how machine learning algorithms are used to identify chest diseases from X-ray pictures. It emphasizes the importance of accurate and timely diagnosis of chest diseases, while acknowledging the challenges faced by conventional diagnostic methods. The review highlights the potential of machine learning as a promising approach to enhance diagnostic accuracy and efficiency in this field. It explores the utilization of both handcrafted features and deep learning-based approaches for extracting informative features from chest X-rays. Furthermore, it discusses the adoption of various machine learning algorithms, including support vector machines, random forests, and convolutional neural networks, for effective detection of chest diseases. The review paper also recognizes the significance of training and validation strategies in ensuring the robust development of models. Additionally, it addresses the potential impact of the proposed methodology on chest disease detection and patient outcomes. The review paper discusses challenges and limitations associated with this approach, such as data availability, interpretability, and ethical considerations. In summary, this extensive analysis strives to make a valuable contribution to the field by offering insights into cutting-edge machine learning techniques, recognizing existing challenges, and proposing potential avenues for future research. [ABSTRACT FROM AUTHOR]

Convolutional Neural Networks (CNN's) are known to perform well on computer vision tasks such as image classification, image segmentation, and object detection. However, one major drawback of CNN's is the huge amount of computing and memory resources needed to train them. In this paper, we propose an architectural unit which we call Upsampling-Based Wavelet Residual Block (UBWRB), that utilizes the 2D discrete wavelet transform coupled with upsampling operators and a residual connection to extract features from image data while having relatively fewer trainable parameters as compared to traditional convolutional layers. The discrete wavelet transform is a family of transforms that find extensive applications in signal processing and time-frequency analysis. For this paper, we use the filter-bank implementation of the discrete wavelet transform, allowing it to act in a similar fashion to a convolutional layer with fixed kernel weights. We demonstrate the performance and parameter-efficiency of CNN's with UBWRB's in the task of image classification by training them on the MNIST, Fashion-MNIST, and CIFAR-10 datasets. Our best-performing models achieve a test accuracy of 99.34% on the MNIST dataset while having less than 120,000 trainable parameters, and 92.90% and 84.27% on the Fashion-MNIST and CIFAR-10 datasets respectively, with both having less than 180,000 trainable parameters. [ABSTRACT FROM AUTHOR]

Natural disasters cause devastation, chaos, destruction, death, displacement, and much more when they occur. With the advancement of recent technologies especially in artificial intelligence and deep learning, the severity and impact of natural disasters could be reduced by predicting their occurrence. In this research, the flood detection (using images) system is developed using the proposed architecture of the Deep Convolutional Neural Network (CNN). In this paper, the main aim is to propose an automated System for precision disaster detection, in particular flood detection. The intelligent system will monitor vast problematic areas that are dubbed flood risk areas during the season. The intelligent automated system will systematically take images with real-time processing to detect floods. If a flood is detected, the system will automatically communicate to the ground station with the early warning to issue an early warning to all the inhabitants in the path of the flood. Since this paper will prove a concept, the other main contribution of this paper is a new dataset consisting of 9000 images collected from various sources on the Internet, labeled, and preprocessed. The Deep CNN architectures were trained and tested by constructing a new dataset of more than 9000 labeled images with flood and non-flood collected from different sources. Several experiments are performed with changing CNN architectural design parameters in order to get the best recognition rates. Experimental results show that 92.50% accuracy was achieved using GoogleNet. The second highest performance was achieved using AlexNet with an accuracy of 92.20%. [ABSTRACT FROM AUTHOR]

Human activity assertion in video is perhaps the most overall applied subjects in the field of picture and video arranging, with different applications in perception (security, sports, and so on), movement ID, video-content-based seeing, man-machine correspondence, and flourishing/handicap care. Development assertion is an amazing measure that a few difficulties. In this assessment, we propose a novel human development insistence methodology utilizing convolutional neural networks (CNN) and perceive the activity performed on that outline annals. Properly, this examination paper will give the required inspiration to seeing human activity possibly reliably (future work). This paper rotates around certification of direct Human improvement utilizing picture dealing with frameworks. [ABSTRACT FROM AUTHOR]

As part of computer vision theory, study and application, object detection and facial recognition continue to be important. In recent years, Deep Learning and specifically Convolutional Neural Networks (CNN) have elicited a great deal of advancement. Also, has consequently gotten a lot of consideration on the worldwide phase of examination about PC vision. The Internet of things assigns actual articles that are inserted with sensors, regulation capacity, programming, and different advancements that associate and trade information with different gadgets and frameworks over the Internet or different foundations organizations. Applying profound brain organizations to IOT gadgets could achieve an age of requesting equipped for performing complex detecting and acknowledgment errands to help another domain of collaborations among people and their actual airs. We have involved a high-level processor Raspberry pi for viable speed and remote capacities. This paper includes the plan of elements for depicting the article and facial qualities followed by reconciliation with classifiers. Haar Cascade classifier is a successful for articles and face identification method. It is an item openness calculation used to distinguish faces in a picture or a constant video. [ABSTRACT FROM AUTHOR]

A major cause of vision loss for humans in worldwide is diabetic retinopathy (DR). This disease should be treated in earlier to reduce the defect. However low-cost settings some ophthalmologists are available to care patients with diabetes, the clinical diagnosis of diabetic retinopathy will a great challenge for doctors. This paper reviews the recent studies on the detection of diabetic retinopathy by using most efficient algorithms of deep learning that is convolutional neural network (CNN) and ALEXNET which is highly used to detect diabetic retinopathy feature from the pictures. CNN and ALEXNET approach to diabetic retinopathy detection which will save time, expensive. And also, it is more efficient. Hence it is more beneficial for diabetic retinopathy detection. [ABSTRACT FROM AUTHOR]

Arrhythmia is an abnormality in the rhythm of the heartbeat. The detection of cardiac arrhythmia in minimal time is important to prevent sudden deaths. Cardiologists use Electrocardiogram (ECG) to detect such abnormalities in the rhythm of the heartbeat. Deep Learning is one of the emerging technologies in classification. We can use deep learning to detect and classify ECG, which plays an essential role in patient monitoring. This paper gives a detailed review of the recent deep learning methods used in ECG for classification. This paper also discusses various deep learning methods such as Convolutional Neural Network (CNN), Deep Belief Network (DBN), Recurrent Neural Network (RNN), Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), and Generative Adversarial Networks (GANs). Among these methods, Convolutional Neural Network is the most suitable method for feature extraction. Classifications based on Deep learning methods are known to be highly accurate. [ABSTRACT FROM AUTHOR]

Face recognition and sentiment analysis used in many surveillance and security application such as biometric identifications, health care applications, and criminal detection. Face recognition framework provide the concept of recognizing the respective face image features to value in the digital forms via using various classification techniques, In this paper, we present the face recognition and sentiment analysis by using the convolution neural network for improving the accuracy of the sentiment analysis system, sentiment analysis covers various gestures such as the happy, sad, neutral, surprise, angry and fear, here our present model based on the convolution neural network and provide the accuracy for the sentiment analysis. [ABSTRACT FROM AUTHOR]

Computer vision is an area of machine learning that enables computers and systems to take actions and make recommendations based on information from digital images, videos, and other visual inputs. The purpose of this paper is to discuss the selection of optimal parameters in car classification training. This paper discusses computer vision training using the CNN algorithm on vehicle miniature classification. One of the factors that affect the performance of Convolutional Neural Network (CNN) is the selection of appropriate parameters. The data we use comes from the primary data we collect. We took photos of 3 types of vehicles, namely buses, cars, and motorcycles with a total of 486 instances. Next we do data preprocessing by resizing 256*256px. Then we conducted various trials of data augmentation i.e: including "horizontal", "vertical", "horizontal and vertical". We experimented with different augmentations to get the highest accuracy. We present several variations of the epoch to identify the best model of CNN. The accuracy results obtained are 97.34%. [ABSTRACT FROM AUTHOR]

The paper presents a study in which the author continues to develop PC software for the translation system to translate specialized texts from English into Russian and vice versa in the construction area. The basis for automatic analysis of a sentence and representation of its structure is graph theory. At the first stage, the sentence is represented by a linear graph. The article discusses ways of assigning features to each word in accordance with its belonging to different parts of speech for their subsequent processing by means of graph convolutional neural networks (GCN). The initial features of each element of the sentence are compared with obtained values , and using certain parameters for the compiled vectors, they are classified as parts of speech. The paper includes parts of the software program for the computer implementation of the proposed machine translation technology. [ABSTRACT FROM AUTHOR]

The study aims to build an efficient feedback system for automated and unmanned restaurants using customers facial expressions by training the model with six different expressions and produce accurate results. The proposed system uses an improved support vector machine for recognizing various expressions of human face in the unmanned restaurants based on a gradient boosting approach. FER2013 dataset has been used for the implementation. This study consists of two groups and the sample size for each group is 392. This research study adapts a novel gradient boosting with kernel trick approach for efficient classification. It is observed that the proposed system obtained 82.14% accuracy while comparing with the existing system which has only 63.88% of accuracy. The result proves that the improved support vector machine with a gradient boosting approach for the feedback system of unmanned and automated restaurants obtained better accuracy than the existing system. This feedback system can also be used as a resort feedback system in order to improve the customer experience. [ABSTRACT FROM AUTHOR]

In many fields, Convolutional neural networks have been efficiently used and by giving more attempts in the area of Seismic imaging. Seismic image analysis is important in a variety of industrial applications like finding underground salt bodies plays a vital role in detecting oil and gas reservoirs. Seismic image analysis still requires experts to examine the Salt body, which is a time taking process. In this paper, we present a method for detecting and optimizing deep-supervised edges that accurately segments the salt body. We create an edge-prediction branch to detect the salt body's border, which aids feature learning by supervising boundary loss. We also used a Sigmoid function (Logistic) which gives output effectively and accurately. We have compared our model with two existing approaches which ended up with good results. [ABSTRACT FROM AUTHOR]

The technique of estimating the complex frequency components of a signal in the presence of noise based on assumptions about the number of components is known as frequency estimation. The problem of frequency estimation appears when we receive a signal from a transmitter with a known frequency, and it is often in radar, sonar, etc. applications. This problem has been dealt with by various techniques and algorithms. This research shows the methods used by researchers to estimate the frequency, This paper discusses the problem of estimating the frequency between a fixed sender and receiver without movement through simulation in Matlab and using two algorithms: the first is correlation method, and the second is the Recursive least squares algorithm(RLS). The best results were obtained by changing the controller gain and filter coefficent from the internal source to the external source. Also, the transfer function was changed to several values to get the frequency response and its better estimate by changing the parameter of the transfer function.the frequency estimation is also discussed in the case of fixed transmitter and moving receiver,which were found in the radar systems with moving targets at a fixed or variable speed. Deep learning techniques were used to detect the target and estimate the frequency by using the bellows of a convolutional neural network(CNN), which proved to be more accurate than others. [ABSTRACT FROM AUTHOR]

Driving while drowsy is one of the significant causes of road accidents, and detection of drowsiness of drivers is necessary to minimize such accidents. There are many methods proposed for drowsiness detection. In this paper, we present the details of a drowsiness detection system using deep learning techniques with ECG as its input. Continuous wavelet transforms (CWT) are applied on the input signal, segmented to 1-minute duration, and then it is converted into scalogram images to be fed as input to deep neural networks (DNN). We use pre-trained transfer learning models, AlexNet, ResNet-50, and VGG-19 for this work. The best results were obtained for Reset-50 with an accuracy of 88.31% without dropping the final layers, which is an improvement of 4.5% over the spectrogram-based implementation. This paper evaluates the effectiveness of wavelet-based features, and we note that the time-frequency domain features outperform. To check the robustness of the net-work, a second model is implemented by dropping the final layers, and the validation accuracy of AlexNet is 51.5%, ResNet-50 is 67.2%, and VGG-19 is 81.8%. [ABSTRACT FROM AUTHOR]

Lately, fake images have become very popular, and it is difficult for people to see them. Due to these fake images, many fields like forensics are facing problems; even social media also became a problem because of the fake images. Many forensics people are trying to overcome this problem. As new types of counterfeit images emerge quickly, the ability to customize new types of counterfeit images is an important, and even challenging, task. In this project, we explore the problem and use machine learning and image editing to overcome this problem. In this paper, we are scheming the LBP on machine learning Convolution Neural Network called LBPNET to get non-fiction images. Here we will first take LBP out of the pictures and train descriptive pictures of LBP with the Convolution Neural Network to produce a training model. Whenever we upload a new test image then that test image will be used in the training model to determine whether the test image contains a fake image or a non-fake image. [ABSTRACT FROM AUTHOR]

This paper proposes an additional 'Median Layer' to the existing Convolutional Neural Network (CNN).The work primarily focuses on removing 'High Density Impulse Noise' present in an image. The proposed median layer performs median filtering on all feature channels. The developed network is an end-to-end network that does not require any non-trivial preprocessing tasks. Based on the experiments performed, results show that, by inserting a median layer to a CNN, the proposed architecture has given convincing results, when compared to the existing approaches. [ABSTRACT FROM AUTHOR]

There are many problems related to image processing and analysis that could be solved using convolutional neural networks (CNN). It's easy to implement CNN using one of thousand high-level programming languages. Such CNN will not be fast and energy efficient enough to be used in real-time systems. The good way to solve this problem is to use special hardware accelerators (neuroprocessors). The paper shows that it is possible to reduce the calculation time of the network using neuroprocessors. The implementation of such hardware is quite challenging process that requires specialized knowledge. That's why we need a tool for automated hardware synthesis. The basis of this tool are various optimizations that will allow us to transfer the network to a hardware platform. The optimization mechanisms in existing tools are either very poor or nonexistent. We propose several optimizations on different stages of developing process of target hardware. In the paper the authors describe a CNN model for handwritten digits recognition and show how to reduce the number of neural network parameters without significant accuracy losses. The authors managed to reduce the number of parameters from 644 120 to 31 530 with accuracy loss just about 0.43%, making the CNN suitable for synthesis on dedicated hardware platform. The authors also examined the dependence of the target platform resources on the method of computing the neural network output (sequentially / pipelined / parallel). It was showed that it is possible to decrease computation time in 7 times using fully parallel computations, bit it required in 4-6 times more resources than using sequential calculations. Using the above, as well as many other optimizations, it allows to create a tool for automated synthesis of high-quality hardware accelerators for CNN. The paper also presents the concept of such tool. [ABSTRACT FROM AUTHOR]

Nowadays, regular intake of nutritious foods is essential for maintaining a healthy eating routine and avoiding obesity in the human body.. In this paper, a novel framework dependent on AI that naturally performs exact grouping of food pictures and gauges food credits. In the preparation section of the model system, it proposes a profound learning model that includes a convolutional neural organization that orders food into explicit classifications. The principle reason for the proposed technique is to improve the exactness of the pre-preparing model. The paper plans a model framework dependent on the customer worker model. The customer sends a picture location solicitation and cycles it on the worker side. Support Vector Machine (SVM), Artificial Neural Network (ANN), and Convolution Neural Network (CNN) are the three classifiers that have been modified to analyze the framework's enhanced precision. Exploring various avenues about a variety of food groups, each with a large number of images, using Artificial Intelligence (AI) to achieve higher grouping precision. [ABSTRACT FROM AUTHOR]

The real estate market is becoming one of the most competitive in terms of pricing and same tends to vary significantly based on various factors. This paper focuses on identifying the type of land and estimating the price using convolutional neural network. Deep learning algorithms exhibit marvelous performance over conventional machine learning algorithms identifying the complex patterns. In this paper two Convolutional Neural Network (CNN) models are used. One is self-proposed model and another is ResNet152V2 model. Models are trained using aerial land image dataset. The ResNet152V2model showed high accuracy compared to the self proposed model. This system helps the land owner to acquire some basic essential details about the land. [ABSTRACT FROM AUTHOR]

In today's digital age, social media platforms like Twitter have become a prolific source of real-time information and public sentiment. Analyzing the sentiment expressed in Twitter posts is crucial for understanding public opinion and making data-driven decisions in various domains, including business, politics, and social trends. This paper presents a comprehensive study on the design and analysis of sentiment analysis for Twitter posts, employing a novel hybrid machine learning approach. Traditional sentiment analysis techniques often struggle to capture the nuances and complexities of human language, particularly in the context of short and informal social media posts. To address this challenge, we propose a hybrid machine learning approach that combines the strengths of both deep learning and traditional machine learning algorithms. Our approach leverages deep neural networks, such as Recurrent Neural Networks (RNNs) and Convolutional Neural Networks (CNNs), to automatically learn features from the raw text data. Additionally, it integrates traditional machine learning algorithms, such as Support Vector Machines (SVMs) and Random Forests, to enhance the model's interpretability and generalizability. The key components of our approach include data preprocessing, feature extraction, model training, and evaluation. We preprocess Twitter posts by removing noise, handling emoticons, and tokenizing the text. Feature extraction involves using word embeddings and other linguistic features to represent the text data. The hybrid model is trained on labeled datasets, encompassing a wide range of sentiments, from positive to negative. To evaluate the model's performance, we employ various metrics such as accuracy, precision, recall, and F1-score. Our experimental results on a large dataset of Twitter posts demonstrate that the hybrid machine learning approach outperforms traditional sentiment analysis methods and standalone deep learning models. The hybrid approach effectively captures the subtleties of sentiment in Twitter posts, making it valuable for real-world applications, including brand sentiment analysis, political sentiment tracking, and customer feedback analysis. In conclusion, this research contributes to the field of sentiment analysis by proposing an innovative hybrid machine learning approach tailored to Twitter posts. The results indicate that this approach is effective in extracting sentiment information from social media data, offering a valuable tool for decision-makers and researchers seeking to gain insights from the ever-expanding world of Twitter conversations. [ABSTRACT FROM AUTHOR]

This research aims to improve the quality of life for visually impaired individuals by identifying obstacles in an indoor environment using a solution approach. The approach uses a Convolutional Neural Network (CNN) to extract features and detect objects from real-time video. A head-mounted image acquisition device detects objects and provides audio information to the visually impaired. The system processes live video streams frame-by-frame, processing each frame as separate images. The authors incorporate additional variables into a novel two-step iterative approach to solve complex computer vision problems. The proposed method improves computational efficiency and accuracy, opening up promising avenues for future research. The paper also presents a new mathematical formulation of curve and surface reconstitute algorithms by introducing auxiliary variables. Moving object detection is crucial for intelligent video monitoring systems, as it allows for accurate extraction of foreground objects. This paper aims to detect real moving objects from un-stationary backgrounds, limiting false negatives and achieving maximum application independence. The model and motion of the target objects are assumed to be unknown. Introduce a new mathematical formulation for curve and surface reconstruction algorithms, introducing auxiliary variables and minimizing energy. This approach transforms an implicit data constraint into an explicit convex reconstruction problem, simplifying it. The formulation also allows for more precise parameter settings, ensuring convergence to a minimum. And demonstrate the properties and results of this new auxiliary problem, mainly when the potential is a function of the distance to the closest feature point. [ABSTRACT FROM AUTHOR]

In this paper a face recognition system using OpenCV's Haar Cascade classifier and convolutional neural network (CNN) is used to suggest songs through Spotify web API. The system intends to extract facial expressions of the user, then the trained model classifies the expression into happy, neutral, sad, surprise, angry, concern and disgust then a song is suggested to enlighten their mood. The model discussed in this paper comprises two models, emotion detection model and music recommendation model, which are clubbed together to give the resultant output. Among existing approaches for emotion detection, the Haar cascade classifier is used here. Correlation uncovers proposed framework execution is cutting edge. In numerous boundaries it beats different frameworks. [ABSTRACT FROM AUTHOR]

Key facial features, gender, age and race, has universal practical applicability be it in identifying the customer demographics for retails, marketing or precisely determining an outlaw in surveillance or law enforcement. Attempts have been made to solve the problem of gender, age and race estimation with convolution neural networks for each task separately which is tedious both in time and labour. In this paper, we tackle this problem with a multitasking Convolution Neural Network (CNN) approach to collectively solve the three tasks on two popular benchmark datasets: FairFace and Adience. Also, this paper highlights some critical aspects that influences the learning of such multi-objective models in contrast to the popular beliefs like 'multitasking models are expected to improve their single-task counterpart' and such. The method of combining the losses of the tasks involved also plays a crucial role in the overall performance of the model. Thus losses are assigned weights accordingly. This paper contributes in developing an algorithm Loss-Proportionate Dynamic Loss Average (LP-DWA), an adaptation of the standard Dynamic Loss Average (DWA) algorithm that assigns dynamic weights on-the-fly during the training process. The proposed LP-DWA method is then validated against a popular benchmark dataset for age and gender problem namely, Adience. Experiments results reveal that this scheme achieves competitive performance with some of the best reported results and is able to reduce the loss value further as compared to its parent counterpart. [ABSTRACT FROM AUTHOR]

In this paper, we propose a novel efficient method for generating a paired training set (clean-noisy pair) from unpaired samples to learn a subsequent system for removing artifacts from images. (In this paper, "noise" or "artifact" is an umbrella term for all unwanted pixels in the image, e.g., rain, snow, random noise.) Previous methods used Generative Adversarial Networks (GANs) and offered considerable performance gain on generating paired samples. They learn the distribution of noise/artifact from real noisy images to replicate similar artifact patterns on clean images. However, since the objectives of these methods is creating images with artifacts, there is no guarantee that they do not modify the (semantic) content of the original image, posing a challenge for this task with unpaired samples. To address this, instead of generating the noisy or clean images, we train models to generate the artifacts mask from noisy samples. Adding such a mask to any clean image converts the image to a noisy one without touching the semantic content of the image. Our adversarially trained Convolutional Neural Network (CNN) is supervised by a discriminator that ensures the mask creates images with similar artifacts w.r.t. the original noisy images. Noisy pair of clean images can then be created by simply adding the generated artifact masks. After that, a de-noising encoder-decoder network is trained on generated paired samples. The proposed method is evaluated on standard benchmarks of various tasks such as de-raining, de-snowing, and removing statistical noise. Results confirm that our semi-supervised method outperforms state-of-the-art methods, even though those that are completely supervised. Besides, the proposed method can improve the accuracy of supervised methods by augmenting their paired samples. [ABSTRACT FROM AUTHOR]