Search

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]

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]

Cotton is one of the important cash crops in India grown on more than 9 million hectors. Like other crops this also suffers from diseases but the effect on the yield affects many livelihoods. Thus, it is necessary to detect the disease in early phase so that cure can be taken. This paper proposes a Convolutional Neural Network (CNN) model for detecting diseases in cotton plants. The model implements in proposed work utilize transfer learning technique, leveraging pre-trained model that have been trained on extensive datasets. Specifically, the Resnet152v2 model serves as foundational architecture for the model, enabling accurate classification of distinct objects. In addition to Resnet152V2 base, the proposed work incorporates two supplementary dense neural network layers. These additional layers enhance the model capacity to learn complex pattern and improve its classification capabilities. With integration of transfer learning and inclusion of supplementary dense network layer, the model demonstrates its ability to effectively classify objects. In addition to Resnet152a 98.36% accuracy in distinguishing between healthy and diseased cotton plants. [ABSTRACT FROM AUTHOR]

Orthopedic implant replacement surgery is a procedure that involves the removal of a damaged or diseased joint and the placement of a new, artificial joint. The new joint is made of metal, plastic, or ceramic materials that are designed to mimic the function of the natural joint. Orthopaedic implant replacement surgery is a highly successful procedure that can significantly improve the quality of life for patients who suffer from chronic joint pain. However, as with any surgery, there are risks associated with ortho implant replacement surgery, including infection, blood clots, and nerve damage. This may lead to aseptic loosening which will become the main cause for surgeries to happen for the second time. Therefore, it is crucially important to diagnose it clearly during the pre-operative and post-operative phases. Our main objective is to assist surgeons in identifying these illnesses accurately during the process, to avoid more serious stages. This paper reviews the methods and approaches for Orthopaedic implant replacement surgery using Computer-Aided Design (CAD), Convolutional Neural Networks (CNN), Augmedics, SCORE-CAM and deep learning algorithms. During all the three phases, various technologies are utilized to improve patient wellness. [ABSTRACT FROM AUTHOR]

Polymer microfluidic devices have gained significant traction in various domains, such as disease detection, drug synthesis, and environmental monitoring. These miniaturized platforms offer a range of advantages, including rapid and high-throughput analysis while requiring only small sample volumes. As a result, they have become increasingly popular in these fields due to their efficient and effective capabilities. A machining technique called micro milling is employed in the manufacture of micro components (micro fluidic devices) such as poly methyl methacrylate (PMMA) or polycarbonate (PC). Micro milling has the advantage of being a quicker, more affordable, and more effective method for fabricating more complex structures. PMMA has been used as the substrate in this study for micro milling followed by factor analysis. The objective of this research is to investigate the impact of individual micro-milling parameters on surface quality. This paper includes 450 microscopic images of the micro-milling substrate by different parameters like spindle speed, depth of cut and Surface quality. The microscopic images are divided to test, train and Val dataset, using three datasets and a Convolutional Neural Network (CNN) is designed. [ABSTRACT FROM AUTHOR]

Pose estimation is a fascinating aspect of pattern recognition that is used in a variety of industries, such as technology, healthcare, gaming, etc. The goal of human posture estimation is to foretell the positions of joints and body components in still photos and moving pictures. Given the wide range of uses for such a system, it constitutes one of the most intriguing domains of computer vision research and has had significant growth. The compact convolutional neural networks architecture for human posture estimation proposed in this study, called BlazePose, is intended towards mobile device legitimate prediction. During analysis, the network operates from over 30 fps on something like a Pixel 2 mobile, providing 33 body keypoints for such a specific individual. As a result, it is particularly well-suited for real-time applications like as fitness tracking and action recognition comprehension. This paper makes two major contributions: a novel body posture tracking approach and a lightweight body pose estimation computational model that evaluates parameters utilizing heatmaps and regression. The suggested method has improved estimate accuracy, which is useful when it is used for robotics, augmented reality, sports and fitness, action recognition, motion analysis, and human-computer interface. [ABSTRACT FROM AUTHOR]

Motorcycle accidents increase due to high speed and traffic congestion. Riding motorcycles without wearing helmet is violating the rules and it may endanger the person's life. This paper aims to detect the motorcyclist failing to wear helmet and to recognize the vehicle license plate to display the number. The system is implemented using convention neural network (CNN) with you look only once (YOLO) architecture. [ABSTRACT FROM AUTHOR]

Crop diseases present a substantial threat to worldwide food security, emphasizing the critical importance of timely and precise disease detection to mitigate yield losses and support sustainable agriculture. This study introduces an innovative approach that harnesses the power of Machine Learning (ML) and the Internet of Things (IoT) to revolutionize the detection of crop diseases. Within this framework, IoT sensors strategically deployed across agricultural fields gather real-time environmental data, while high-resolution crop images are captured using drones and on-site cameras. ML models, including Convolutional Neural Networks (CNNs), harness this data to enable early and accurate disease detection. By integrating IoT capabilities, we can correlate disease outbreaks with environmental factors, thereby enhancing the decision-making process. The results are promising, consistently demonstrating high accuracy in disease detection and the potential for reducing the need for chemical treatments. This research paper signifies a significant step towards promoting sustainable farming practices and enhancing global food security. [ABSTRACT FROM AUTHOR]

Traffic forecasting and prediction are crucial in urban planning, transportation management, and decision-making. Traditional methods often struggle to capture traffic data's complex spatial dependencies and temporal dynamics. In recent years, Graph Convolutional Neural Networks (GCNs) have emerged as a powerful tool for addressing these challenges. GCNs leverage graph structures to model relationships between traffic nodes, such as road intersections or segments, and have shown promising results in various traffic forecasting and prediction tasks. Within this paper, an extensive exploration unfolds, delving into the utilization of GCNs in the realm of traffic forecasting and prediction. It meticulously illuminates their inherent strengths, navigates through limitations, and envisions potential future trajectories. Additionally, a pioneering framework for traffic forecasting takes centre stage within these pages. [ABSTRACT FROM AUTHOR]

In many applications like human-computer interaction and video analytics, facial expression analysis has become one of the prime objectives. A simple approach for finding the comfortable level/happy mindset of a human is to recognize their smile. Hence, in this paper, we proposed an automatic smile recognition technique with high-level features extracted using convolutional neural networks. Here cascaded convolutional network model with YOLO V4 and LeNet-based architecture is used to detect the person, and, smile recognition. The experimental results of our proposed system provide higher accuracy of 96.82% than the existing techniques. [ABSTRACT FROM AUTHOR]

This paper's objective is to improve accuracy in plant leaf disease detection using Machine learning techniques such as K-Nearest Neighbors classifier (KNN) with a sample size of 10 and Novel Convolution Neural Network (CNN). This article consists of 2 groupsi.e Convolutional Neural Network (CNN) and K-Nearest Neighbors Classifier with a sample size of 10 for each group. G Power software is used to determine sample size with a pretest power value is 0.8 and alpha 0.05. The accuracy evaluated using both algorithms is compared and CNN appears to be the most significant with an accuracy of 92.48% and KNN with an accuracy of 74.14% with p=0.041. An Independent sample t-test has been used and analyzed with a confidence level of 95 percent. Results show that the CNN algorithm is significantly better than KNN in detecting plant leaf disease. [ABSTRACT FROM AUTHOR]

The Dubas bug (Db) is a deadly pest that affects palm and agricultural crops; however, early automatic identification of frond diseases of this type can thus help reduce human effort and reduce economic losses. Traditional methods of detection rely on human vision and hands to spot and categorise the palms affected, while more recent research has begun applying deep learning networks due to their effectiveness in classification tasks. This paper presents a strategy of automatic recognition of Dubas blight based on identification of diseased palms. Data was collected in natural climatic conditions using a drone camera and several pre-processing steps were applied. In particular, pre-sets were created by slicing the images into small parts and using Lifting Wavelet transformations (LWT) to shrink the data in size; this was then used to train a convolutional neural network (CNN) with the assistance of pre-trained Xception, InceptionV3, DenseNet121, and ResNet101 models with different parameters. Dropout and data augmentation was used to avoid overfitting, and the highest classification accuracy was determined to be that offered by the Xception model, with experimental results suggesting that this developed more than 99% accuracy. [ABSTRACT FROM AUTHOR]

Handwriting is a number, letter, word, or sentence written on a piece of paper that comes from human or individual handwriting. Handwriting has benefits that can help fine motor coordination, memory, and cognitive development in multi-sensory activities. There are several problems in the world of education that involve handwriting. First, handwriting that is difficult to read can cause students to get low grades and even fail to graduate from school. Second, sloppy handwriting makes students lose confidence. Third, students who have bad handwriting have the potential to be bullied by friends at school. The description of these problems causes the author to be interested in developing a digital technology innovation in detecting the quality of handwriting. The purpose of this research is to detect handwriting quality using a Convolutional Neural Network in transfer learning process of EfficientNet B0 model architecture on fine tuning. The data used are secondary data from the CSAFE Handwriting Database at Iowa State University collected by researchers at the Center for Statistics and Applications in Forensic Evidence. This data consists of 27 scanned handwriting samples from each of the 90 participants for a total of 2430 handwritten image samples. The model generated from EfficientNet B0 SGD optimization using learning rate of 0.01 at 28th epochs is very good model obtained. Moreover, the model was evaluated for accuracy, precision, recall, and F1−score of 92%. Next, the model is integrated into cloud computing. Furthermore, implementing the best model, android application is developed called Rayuan (Rate Your Handwriting). [ABSTRACT FROM AUTHOR]

Multi-class image classification is one of the tasks of computer vision in which convolutional neural networks are widely used. In this paper, we create a novel ensemble learner by stacking InceptionV3, Xception, MobileNetV2, and ResNet50 convolutional neural network in an attempt to outperform the best performance of each individual models. In an ensemble learning algorithm, by combining multiple different models with different architectures, an error of prediction from a single model can be corrected by other models, thus reducing the likelihood of selecting a wrong prediction. The dataset that we used to do benchmarking is the Intel Image Classification dataset which consists of 6 classes that displays images of fully colored natural scenes. The training set consists of 14034 images, while the test set consists of 3000 images, all having 150x150 as their default size. The training process on all models involves reducing their learning rate when the performance plateaus, and early stopping if there aren't any significant improvements from an epoch to the next. The result is that each individual models have an average of 83% accuracy, yet our ensemble model managed to reach 88% on the test set. [ABSTRACT FROM AUTHOR]

The reality is that the population and number of automobiles on the street are growing day by day. The efficient monitoring of continuously increasing traffic is a critical challenge that traffic operators are facing. Traffic control is generally achieved using stationary video recorders, observers or manual counters. Hence, Unmanned Aerial Vehicles or drones are the high-quality answers to this trouble as they lessen the price of the set-up of numerous cameras and sensors. This paper's principal ideology is split into phases. The first one is the monitoring of traffic using a drone which helps in analyzing the traffic in a particular area and making interpretations. The second is fire detection and extinguisher. Here the drone is used to detect fire with the aid of a fire sensor as it helps the fire-fighters to analyze the affected area and prioritize their work. This project also up-skills in designing a quad-copter. Our motivation for this project lays both personal interests in a better understanding of object detection and academic research. A drone consists of a camera to cover a wider area which helps in traffic analysis. Here, the number of vehicles are estimated. Drones have recently emerged as an outstanding technology in the business market and are geared up with a camera, geo-positioning sensors, and communications hardware that may relay statistics. This project is implemented by using artificial intelligence and the Internet of Things for Fire Detection. To be more specific traffic detection is achieved using Deep Learning algorithm which is the Convolutional Neural Networks Algorithm using python. Fire sensors are mounted on the drone, and it flies on a flight path within the building and sends fire alert data to the system controller upon sensing a fire condition via fire sensor and also puts off the fire using the water pump. For fire detection the sensor used in the drone basically detects IR (Infra-Red) light wavelength that is emitted from the fire flame which helps fire-fighters to quickly assess and monitor dangerous areas which results in more safety and limits the time it takes to scan the whole area. [ABSTRACT FROM AUTHOR]

This paper investigates the most commonly used Deep Learning (DL) based techniques for estimating the wireless channel at the User Equipment (UE). The generalized Nakagami-m fading distribution is considered for deep learning network. The motive of this work is to provide an experimental investigation on the wireless channels using three different deep-feed forward neural networks, namely Artificial Neural Networks (ANN), Long Short Term Memory (LSTM), and Convolution Neural Networks (CNN). The analysis includes the preprocessing approach for handling complex received signals as one of the inputs and actual channel coefficients as output. Bit Error Rate (BER) performance analysis is performed and is compared with the classical LS estimator. It is demonstrated that DL channel estimation outperforms the Least Square (LS) based estimation, and simulation results support that DL-based channel estimation is very efficient. [ABSTRACT FROM AUTHOR]

This paper explains the training process of the ResNet-50 architecture Convolutional Neural Network (CNN) model from the Keras application and the implementation of the model in the attendance system prototype. This project aims to find the best optimizer function for the face recognition model configuration and evaluate the model with an attendance system prototype. Dataset used in this project is the authors' institution's employees' face image dataset. Six classes with seven images each are collected as a training dataset. Trial-and-error experiments with different model configurations are done to get the best model. The best result is implemented in the attendance system prototype. The model was trained with 500 epochs at the beginning of an experiment and more than 50 epochs if the result is not satisfying, and re-configure learning rate, batch size, and the optimizer function if there is no accuracy improvement after epochs addition. This project concluded that the best optimizer function is Adam, and modification of the CNN model configuration significantly improves accuracy. As a result, in the 4th experiment, ResNet-50 reached 95% accuracy on testing and 100% on training with Adam optimizer function, Softmax activation function as output layer, 16 as the batch size value, and 0.001 as the learning rate value. The accuracy is not stable even until 700 epochs. However, this result achieves better accuracy than the previous project, a face recognition system using Linear Binary Pattern Histogram (LBPH) algorithm done by Sarwar and a face recognition-based attendance system using custom CNN architecture by M. K. Rusia. [ABSTRACT FROM AUTHOR]

Patients with this malignancy have a few decades less time left in their lives. If a brain tumour is discovered too late, the patient may go into a coma or become comatose and die. It is essential to identify at an early stage and segment brain tumour regions in MRI In order to perform clinical assessments such as, diagnosis, tumour detection treatment planning, and evaluation,. For segmenting brain tumours in MRI images, a better encoder-decoder convolutional neural network (CNN) architecture is proposed in this paper. Three encoding processing blocks are initially used to extract the hierarchy of tumor tissues characteristics from each input slice. Then, in the following three decoder blocks, these features are enhanced to obtain the original image size. Finally, the extracted features are classified using softmax classification. Concatenated feature maps are used in the decoding stages of this encoder-decoder CNN-based segmentation method, which calls for more processing power. For the BRATS-2018 dataset this technique obtains 91%, 93%, and 94% of dice score values. [ABSTRACT FROM AUTHOR]

Technologies focusing on the visually impaired play an important role in assisting the visually challenged and easing their living standards. Previously there have been several other technologies and papers proposed under this category to help the visually challenged. Applications have been built, and are being developed, to help these people work independently without anybody else's need and complete daily tasks easily, without any difficulty using Artificial Intelligence. The proposed application, Noon Gil, supports voice commands and performs operations accordingly using architectures and algorithms like The technologies used are Optical Character Recognition (OCR), Convolutional Neural Networks (CNNs), Natural Language Processing (NLP), Mobile Vision API, TensorFlow Lite, Google ML Kit, MobileNet SSD, etc. Our project which is an android application with an overall accuracy of 94.74% comprises three main functions: Text Detection with an accuracy of 90.83%, Object Detection with 94.40% accuracy and barcode scanning with 99%. [ABSTRACT FROM AUTHOR]

Sign language is a crucial mode of communication for individuals who are deaf or hard-of-hearing, and a reliable sign language recognition system can greatly aid in bridging the communication gap. According to statistics provided by the Government of India, there are only around 250 accredited sign language interpreters in India, servicing a deaf population of 1.8 million to 7 million individuals. This paper presents a real-time sign language recognition system for Indian Sign Language (ISL) using YOLOv8 for object detection. The system uses a single convolutional neural network (CNN) architecture to process the entire image in one forward pass and predict the bounding boxes and class probabilities for each sign in the image. The results of the study demonstrate the feasibility and potential of the system in revolutionizing communication for the deaf and hard of hearing community and have applications in gesture-controlled robotics, home automation, game control, human-computer interaction, and sign language interpretation. [ABSTRACT FROM AUTHOR]

With the upcoming large-scale surveys like LSST, we expect to find approximately 105 strong gravitational lenses among data of many orders of magnitude larger. In this scenario, the usage of non-automated techniques is too time-consuming and hence impractical for science. For this reason, machine learning techniques started becoming an alternative to previous methods. In our previous work [1], we have proposed a new machine learning architecture based on the principle of self-attention, trained to find strong gravitational lenses on simulated data from the Bologna Lens Challenge. Self-attention-based models have clear advantages compared to simpler CNNs and highly competing performance in comparison to the current state-of-art CNN models. We apply the proposed model to the Kilo Degree Survey, identifying some new strong lens candidates. However, these have been identified among a plethora of false positives, which made the application of this model not so advantageous. Therefore, throughout this paper, we investigate the pitfalls of this approach, and possible solutions, such as transfer learning, are proposed. [ABSTRACT FROM AUTHOR]

Cervical cancer is one of the most known health problems faced by women around the world. Early detection of cervical cancer may reduce the mortality rate. Pap smear images are new techniques used for screening cervical cancer. This paper excludes the nucleus of pap smear images, and the resultant images are classified into seven classes based on the surrounding region nucleus. Automated features are extracted using three pre-trained convolutional neural networks (CNN). The resultant features are twenty-one. The principal component analysis reduces the dimensionality and selects the most significant features into ten features. These features are fed to two types of machine learning algorithms: support vector machine (SVM) classifier and random forest classifier. The support vector machine classifier achieved the highest accuracy for seven classes, reaching 93.1%. This method will help the physicians in the diagnosis of cervical cancer depending on the tissues, not the nucleus. Furthermore, the result can be enhanced using a huge amount of data. [ABSTRACT FROM AUTHOR]

Face recognition is a biometric based application software capable of verifying or identifying exclusively a person by analyzing and comparing person's facial contours using patterns. The main applications of face recognition system is found in security systems, law enforcements and other initiatives. The variation in illumination, effects immensely on recognizing faces or images. Even for the familiar faces, illumination variations lead to alteration in the shadow shapes, shift in locat ions, fluctuations in features and contract gradient variations. Under these conditions, larger impact is observed to recognize these face images. A face recognition system is developed under these varying illumination conditions in this paper. Various methods of illumination normalization like Self-Quotient Image (SQI), Histogram equalization (HE), Gross & Barovick's anisotropic smoothing (GB), Gross & Barovick's anisotropic smoothing (GB) and Multiscale Retinex (MSR) are comparatively analyzed. Illumination invariant face recognition is developed by first acquiring 40 illumination variants face s. Face detection is performed by deformable template with scaled color map and its negative. Further face alignment is estimated by supervised learning by estimating the locations of a set of facial landmarks. Later image normalization is performed using Tan Trigg's Normalization Technique (TTVT) which includes Gamma correction, MaskingDifference of Gaussian (DOG) and contrast stretching methods [1]. The resultant normalized face image is classified using Convolutional Neural Networks. The evaluation and performance analysis are discussed to determine the applicability and suitability for various applications. [ABSTRACT FROM AUTHOR]

This paper proposes and implements a real-time word-level Arabic Sign Language (ARSL) recognition system. The work aims to facilitate the interaction between people with hearing loss and the general public. The system uses a frequency-modulated continuous wave (FM-CW) radar operating at 77 GHz millimeter frequency to transmit and capture signals from a person performing hand gestures in front of the radar. The signals are then used to train a convolutional neural network (CNN) to translate the gestures into words. The CNN is built using the Keras platform on Python and has ten layers, including two 2D convolutional layers. In the preliminary work, the system is designed to identify ten important common words. Preliminary results show that the system can classify gestures correctly in real time with 81% accuracy. [ABSTRACT FROM AUTHOR]

Tampered Medical images and scans are a very serious issue in the medical field. The results of tampered scans can range from mild to serious results. For example, a tampered scan or medical image can lead to misdiagnosing a patient with a serious condition or a patient who suffers from a serious condition can be misdiagnosed as healthy leading to a delay in receiving proper treatment. In this paper, we propose a deep learning-based methodology to detect tampered/fake cancers in 3D CT scans of human lungs. We use a publicly available dataset of deepfakes that include both tampered and genuine cancers and propose the use of a modified Convolutional Neural Network (CNN), in particular the AlexNet with Transfer Learning. The model was pre-trained on a large dataset and fine-tuned on the medical images dataset to detect tampering. Experimental results achieve a high level of accuracy in detecting tampering. Using the AlexNet with Transfer learning, we achieved an accuracy of 89.47%, Recall of 89.47%, Precision of 89.47%, and F1 measure of 89.47%. This is higher accuracy than similar methods using the same dataset reported in the extant literature. Achieving this high accuracy for a multiclass complex problem is considered an excellent achievement. Expert radiologists are mostly unable to distinguish between real and tampered cancer images. The results demonstrate the potential for deep learning-based models in improving accuracy and efficiency in tampered medical scan detection. [ABSTRACT FROM AUTHOR]

A pothole is the most prevalent sort of road issue that must be assessed before repairs can be carried out. Along with the regular wear and tear of vehicles, it is a major factor in traffic accidents. The collecting and analysis of data on road defects is vital for the evaluation of road concerns. Data collection through a variety of imaging modalities has been automated, but fault assessment is still performed by hand. Slowing down the highway maintenance process by manually locating and grading potholes is costly and time-consuming. As explained in this paper, potholes in road images can be detected and classified using convolutional neural networks. In order to classify images into potholes and non-potholes, we used convolutional neural networks and trained models. As opposed to seven pre-trained models, convolutional neural networks were compared to see how well they performed in terms of precision and recall accuracy. Potholes may be accurately detected on road photos by trained models InseptionResNetV2 and DenseNet201 with an accuracy of 89.66 percent, according to the research findings. [ABSTRACT FROM AUTHOR]

Due to the fast spread of the new Coronavirus called SARS-CoV-2 (otherwise known as COVID-19 virus) worldwide and its continuous mutations, such large viral outbreaks require early elucidation to determine the genetic sequence of the virus to design an effective system for identifying variants Different known and unknown virus. Identifying specific variables facilitates the comprehension and modelling of viral propagation patterns to treat and manage outbreaks, create efficient reduction techniques, and avoid future outbreaks. In addition, it serves a vital role in assessing the efficiency of known vaccines against each variation and modelling the probability of superinfection. It is known that the genomic sequence carries the vast majority of genetic information pertaining to variations and variants of the Coronavirus. This paper proposes a system that depends on deep learning to predict the mutations of complete COVID-19 virus genomes based on the Convolutional Neural Network (CNN) algorithm as an alignment-free method. The k-mer technology is applied to fragment the DNA of coronavirus mutants to create a unique vocabulary. The proposed approach can correctly predict and characterize other coronavirus strains, such as MERS-CoV, Alpha-CoV, SARS-CoV-2, SARS-CoV, Gamma, Beta, and other strains, regardless of missing information and sequencing errors with an accuracy rate of 99% on the data set. [ABSTRACT FROM AUTHOR]

After a deep questioner among students, faculty members and higher education experts. The demand for qualified engineers in the specific field as understandably gone down. The situation is grimmer for tier-2 and tier-3 colleges in India. The purpose of opting engineering course is not to get expertise in the particular course, instead to get a job in government sector with valid degree. The problem identified in the field of engineering education towards improving the quality in education is addressed with the help of exploratory data analysis. The dataset used in our experiment is collect from UCI machine Learning Repository having 33 variables and 1044 observations. The contribution made in the paper is to identify the vital attributes using single and multi-variant regression techniques. To perform prediction using Decision Tree, Random Forest, Support vector Machine and compare their performance in terms of classification accuracy and F-Score. In addition to that a convolution Neural Network (CNN) model is established in which, the vital attributes identified using regression techniques are provided as inputs and weights at each stage is estimated using Gradient Decent algorithm with step size 0.5. The classification accuracy from 63% is improved to 97% with the help of CNN model in 26664 iterations. The finding of the present research states that student having backlogs are less frequently opting for higher studies. [ABSTRACT FROM AUTHOR]

Detection of diseases in crops is an important aspect to deal with as it affects the productivity of the agricultural yield. The farmer may not realize that the crop is suffering from a disease by looking at the leaves. In this paper, we will be performing Convolutional Neural Networks (CNN) where data needs to be pre-processed and then we need to divide the data into training dataset and test dataset. We will use the tomato leaves dataset to predict whether a tomato plant is healthy or diseased. Finally, we will be giving a user input image and using this model we will be detecting whether the plant is healthy or diseased and which disease it is suffering from which will help in enhancing the life of a farmer by increasing his productivity. Based on this detected information the farmer will be cautious if the plant is diseased and take appropriate actions. [ABSTRACT FROM AUTHOR]

Feeling might be a hot subject in a very kind of regions, just as medication designing, brain research, neurobiology, and wellness. Feeling recognition may help inside the diagnosing of mental and social issues. Profound learning has made significant advances inside the space of picture acknowledgment lately. during this paper, we have an inclination to propose a LeNet configuration upheld Convolutional Neural Networks (CNNs) for biometric verification. first of all, we have an inclination to coordinated three data sets (KDEF, JAFFE, and our custom data set) (JAFFE, KDEF and our custom dataset). At that point we have a propensity to prepared our LeNet plan for feeling states grouping. during this examination, we have an inclination to accomplished a characterization exactness of ninety six.43 p.c and an approval precision of ninety one.81 p.c exploitation looks to order seven totally various feelings. [ABSTRACT FROM AUTHOR]

Potholes have proven to be a major roadblock in recent times. Potholes are one of the leading causes of car accidents and other untoward incidents on the road, and they also cause vehicle wear and tear. The collected data is currently being processed in order to determine road faults and work on them. Data collection is almost fully automated thanks to a variety of imaging systems, but the evaluation of defects based on the collected data is still performed manually and is not automated even today. This method of classifying and evaluating potholes is expensive, time-consuming, labor-intensive, and repetitive, and inevitably, the total road maintenance procedure is slowed. We have described a new detection technique for potholes in this paper. Convolutional neural networks (CNN) have been used. Using this approach in the proposed system, input images were classified into two categories. Implementation was done using OpenCV library in Python. It was trained on 722 raw images and tested on 198 images, with precision, correctness, and recall metrics among the metrics used to assess the results. [ABSTRACT FROM AUTHOR]

The deaths caused by road accidents are one of India's most pressing issues. The inattentiveness of the driver is responsible for almost 80% of all collisions. Mobile phone use, conversing with passengers, reaching behind the wheel to grab something, and drinking while driving are just a few of the factors that can cause a driver to become distracted. This project aims to create a reliable and effective system for detecting and classifying distracted drivers. The efficacy of CNNs in the field of computer vision was the inspiration for this paper, which proposes a paradigm based on Convolutional Neural Network which not only senses but also recognizes the driver who is distracted. For this purpose it is using a publicly available dataset from the Kaggle website as an input for this model. There are several different forms of distractions, but this study focuses on the manual distraction, which is focused on the driver's stance. The goal of this work is to keep an eye on the driver's actions and seek for diversion signs. Distractions are divided into two groups in the model: safe driving and unsafe or distracted driving. For this purpose, various architectures of CNN can be used. This work, however, employs sequential architecture and achieved the highest observed accuracy of 99% with a total of 9.5M trainable parameters. The primary goal of this research is to classify a test image into two states of the driver that have been considered. [ABSTRACT FROM AUTHOR]

Aim: The main aim of this paper is to recognize mango leaf disease using convolutional neural networks in google co-laboratory. Materials and Methods: A dataset of 120 photographs of diseased and normal mango leaves was used to distinguish mango leaves using the CNN technique which is based on a deep learning algorithm. Results: The proposed 5D CNN model has a 98.67 % accuracy and 2D CNN has a 89.05% accuracy in leaf disease classification, so the novel 5D Convolution Neural Network algorithm appears to be better than 2D CNN standard deviation is 1. Conclusion: In mango leaf disease detection, the proposed novel 5D Convolution Neural Network based leaf disease recognition model shows less training time and high training accuracy. This demonstrates the adaptability of CNN models in real-time implementations. [ABSTRACT FROM AUTHOR]

The paper aimed to detect the brain tumor with MRI images through Novel CNN and predict the accuracy value. Materials and Methods: Proposed work involves two groups for brain tumor detection. Group 1 was the Novel Convolutional Neural Network (CNN) algorithm and Group 2 was Decision Tree (DT) algorithm. The sample size was calculated. It was identified that 10 samples/group and 20 samples were taken totally. The improved CNN algorithm was compared with the DT algorithm for predicting the accuracy. Results and Discussion: Data collection was carried out and the analysis was done in the Google Collab for the execution of results and for measuring the accuracy of given algorithms. The Novel CNN algorithm produces high accuracy with 92% whereas the DT algorithm gives 91% of accuracy. In this work, the improved CNN algorithm was compared with the existing DT Algorithm. Conclusion: The data was collected from various resources for the usage of brain tumor detection. The Novel CNN algorithm obtains high accuracy when analyzed with other existing algorithms. [ABSTRACT FROM AUTHOR]

Pothole detection is a challenging task as it involves inhomogeneity in intensity, complex topologies, contrast less, and noise rich background. Still, it's a very important task concerning the maintenance of roads. Potholes are being detected using computer vision which generally involves analysis of a two-dimensional image of the road or modeling of the road surface. Provided both these methods are independently used. The accuracy of this method of pothole detection is far from the top point. These bring the usage of digital image processing which develops the detection of potholes with at most accuracy. Digital Image processing provides strong receptivity, usage of huge actions, and broad benefits. In this paper, we propose a system that supports digital image processing techniques of high efficiency to result in consistent and accurate detection of potholes. This project studies an intelligent digital industrial pothole detection mechanism to increase human safety, detection capability and decrease the factor of risk. Firstly, the sources of images from the road are pre-processed to remove noise and other irrelevant features. Then a training model was developed to train the system to classify between path holes and normal cracks using convolution neural networks and MATLAB. Also, an IoT device is used in this system which updates the detected potholes on the road to the respective web server along with the location coordinates, as a part to notify a complaint to check. This system works on providing a better pothole detection technique and also updating the same to the concerned. [ABSTRACT FROM AUTHOR]

Medical imaging has seen substantial improvements not only in image procurement but also in terms of interpretational methodologies. The practice of using technologies such as image processing and artificial intelligence has shown remarkable results in the detection, classification, and localization of human bone fractures. Breakage or cracks in the bone occur when an external force is applied to a bone that is significantly larger than the bone's tolerance, which may lead to dislocation requiring extensive treatment. However, manual examination by experts is a complex and arduous process with a high chance of false detection. Many approaches under Artificial Intelligence and its subsets are proposed to facilitate fracture diagnosis and prevent misinterpretation, including Deep Learning and Convolutional Neural Networks (CNN). This paper is a comprehensive study of the various techniques and models designed using image processing approaches and the advances in machine learning and Deep Learning that have demonstrated improved performance in the precise detection of fractures in human bones. [ABSTRACT FROM AUTHOR]

In this paper, Convolutional Neural Network Segmentation (CNNS) by Dual-Interface Reign Architecture is proposed to figure out the condition of breast tumor. The existing novel approaches from machine learning may not be voluntarily useful as they require a huge training data bit with segmented pixels. This architecture is built upon a CNN and a dual mode network which is used to segment and can non-linearly maps the input data into a deep latent space respectively. the two paths of dual-interface networks are locality preserving learner and a conditional graph learner, both are used for the extraction of images and exploit the input features. The complemented dual-interface architecture is further used to improve the rate of classification. This proposed method can overwhelm the restriction of some benchmarks for the classification in breast imageries. In addition, to perform extensive error analysis, an analytical model is developed for mammograms and tested for varying pixel widths and input probabilities. The analytical model is validated through simulation. The results of our proposed method demonstrate that it can effectively detect the different kinds of mammograms in database by segmenting them more precisely. When comparing with previous works the precision is improved up to 5% and 8% than the GSA-ANN and ANN respectively. [ABSTRACT FROM AUTHOR]

Picture captioning has recently become a new difficult challenge that gathered everyone interest, which is capable to define automatically an image's content with properlyformatted text English sentences. it makes a great impact by assisting for visually impaired people better recognition of their circumstances. By taking the images of neighboring environment that make use of these photos to generate captions that are examined visual amplification impaired, so that they can get a better sense of what's going on around them. In this paper to extract features, a mixture of convolutional neural networks of the images & then LSTM was used (Long short-term memorys) to regenerate text from these features. The obtained text is converted into speech so that it can be read out. Our model generates highly descriptive captions that canpotentially greatly improve the lives of visually impaired people. [ABSTRACT FROM AUTHOR]