Search

In the present day, deep learning applications are widely used in every field of development. In this paper, such a type of application is discussed in real-time. Mura dataset is the largest abnormal dataset given by the University of Stanford. In this research paper, we will understand the pre-processing of the image dataset. Kera's and TensorFlow libraries are used to build this model. We are classifying images using different ML models where accuracy is compared to get the best model build. This paper gives us a clear understanding of building deep learning models, understanding pre-processing techniques, and working on heavy data. This research's main agenda is to create a model which can classify bone fracture images and also give a small path for further development in the deep learning area. In this project we are using Resnet50, InceptionV3, Mobile Net and DenseNet with optimizer Adam and comparing their accuracies graphically. [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]

Monkeypox disease was declared a Public Health Emergency of International Concern (PHEIC) by the World Health Organization (WHO) on July 23rd, 2022. The purpose of this designation is to encourage the global community to take coordinated measures to contain the spread of the disease and protect communities. The recovery period for Monkeypox disease typically ranges from 2 to 4 weeks, and in some cases, the disease can resolve spontaneously with or without treatment. However, severe cases have been reported, which can lead to fatalities. The fatality ratio for Monkeypox disease is estimated between 3% to 6% at present. It is essential to take precautionary measures to prevent the spread of this disease and seek medical attention if any symptoms are observed. Monkeypox is not highly contagious, and it generally spreads through close contact with an infected person. However, in May 2022, a significant number of cases of Monkeypox disease were identified, indicating that the disease was spreading rapidly. In light of this situation, it is crucial to track and isolate suspected cases promptly to prevent further transmission. Rapid and effective contact tracing, along with timely isolation of cases, is essential to contain the spread of the disease and prevent it from becoming an epidemic. Identifying, controlling, and taking measures to control the spread of Monkeypox disease is of utmost importance. To aid in this effort, this paper aims to create a large-scale dataset of Monkeypox disease images using Generative Adversarial Networks (GANs). GANs have proven to be highly effective in generating high-quality images and can aid in creating a more comprehensive and diverse dataset. The generated images will be used to train different deep learning models for image classification to diagnose Monkeypox disease. The deep learning models will be compared, and their performance will be evaluated using images of infected human skin with Monkeypox disease and healthy skin images. The results obtained from this paper will improve the accuracy of previous works, aiding in the early diagnosis and treatment of Monkeypox disease. [ABSTRACT FROM AUTHOR]

This paper reviewed the Hand Sign Recognition system from 2018 to 2022. Based on the review, much research is done on Deep learning-based hand sign recognition. Many of these studies are published annually in article, journals, and conference proceedings. Additionally, we also examined the performance of the HGR system in terms of recognition accuracy. In this paper we concentrate on the data environment, the data acquisition, and the hand gesture presentation— three crucial components of the vision-based HGR system. We also looked at how well the vision-based HGR system performed regarding recognition accuracy. [ABSTRACT FROM AUTHOR]

Various machine learning algorithms are developed with an aim to create precise and trustworthy models and extract knowledge from data sources. Deep expertise in the field of machine learning is required for the challenging task of choosing the right algorithms for a specific dataset. There is no single algorithm that outperforms all others across all applications and different datasets. The difficulty of choosing an appropriate algorithm for a specific task in specific domain is related to the properties of the dataset. Properties of the dataset are measured through meta-features. Meta-features describe task and can provide explanation how one machine learning approach outperforms other algorithms on a given dataset. Learning about the effectiveness of learning algorithms, or meta-learning was developed to deal with this issue. Focus is required because previous research papers have not successfully identified meta-features in particular domains. In this research, we have evaluated various meta-feature characterization methodologies and have concentrated on basic meta-features. Business domain data is in the focus of this paper. We computed basic (general) meta-features and illustrated several use cases for their applications. [ABSTRACT FROM AUTHOR]

The paper proposes a novel approach for recognizing and classifying gym exercises using human pose detection with Mediapipe and LSTM, followed by matching the accuracy of the exercises using the Hungarian algorithm and cosine similarity. The proposed approach involves the use of a combination of cutting-edge technologies, including computer vision and deep learning, to recognize and classify a range of gym exercises accurately and efficiently. This study presents several advantages of this approach, including the ability to leverage temporal information, robustness to variations in lighting, camera angle, and occlusions, and the ability to handle a wide variety of gym exercises. Experimental results demonstrate that the proposed approach achieves high accuracy and outperforms existing methods for gym exercise recognition and classification. The paper also compares the proposed approach with existing techniques, highlighting its strengths and cost-effectiveness. It concludes by outlining the methodology, including dataset collection and pre-processing, landmark extraction, LSTM model training, exercise classification, and similarity analysis. Finally, the paper discusses potential future research areas in this field. [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]

Road traffic by vehicles is continuously increasing through the world and they lead to terrible traffic jamming and significant delay in emergency services at cross roads. The vehicles stuck in that traffic jam may reach their destination with more delay which was lead to loss of lives or a property in some cases for example, emergency vehicles when stuck in the traffic jam like ambulance, fire engine vehicles, police jeeps, etc. Therefore it becomes most significant to control the vehicle traffic on the roads so as to acquire the effective control over the vehicular traffic based on the priority that may leads to saving the life of people in emergency situations by providing delay less services. This paper presents a survey on various methods and systems developed for the traffic classification according to the priority of emergency vehicle and for scheduling the vehicles accordingly in traffic. In this paper, different deep learning methods explored by various works are presented for effective increasing of vehicles creating traffic on the route of emergency vehicle. Finally, the works presented in this paper are compared and result analysis can be illustrated. [ABSTRACT FROM AUTHOR]

The paper presented the various techniques that can be applied for vehicle number plate identification. The extensive literature survey has been carried out on already presented technologies. Out of these technologies, three technologies have been selected and explained in this paper, the technologies are: OCR Method, Deep Learning Method and OpenCV Method. The comparison of three different technologies also available for vehicle number plates identification has been carried out on the basis of accuracy, complexity and cost. [ABSTRACT FROM AUTHOR]

Fingerprint is one of the most famous biometric characteristics. It is employed in many fields such as forensic, security, recognition and classification. This paper focuses on clustering fingerprint images into original and fake. Unsupervised Deep Leaning (UDL) is proposed, it exploits the Self-Organization Maps (SOM) to provide such clustering. It consists of two internal processing parts. The first part is for the feature extraction. The second part is for the unsupervised clustering of the SOM. Fingerprint images from the ATVS-FakeFingerprint DataBase (ATVS-FFpDB) for without cooperation are utilized in our work. Multiple clustering and classification metrics of the Silhouette Value (SV), Calinski Harabasz Index (CHI), Davies-Bouldin Index (DBI) and accuracy are provided. Also, different comparisons with state-of-the-art Deep Learning (DL) architectures are provided. Our UDL approach has achieved a high accuracy result of 92.86% and fingerprint images are successfully clustered into original and fake categories. [ABSTRACT FROM AUTHOR]

Blood Brain Barrier (BBB) is a membrane made of a network of tissues which protects the brain from noxious agents that could cause brain infections. It only allows small molecules and lipid-soluble molecules to pass through the membrane. Brain drugs need to penetrate the Blood Brain Barrier (BBB), so molecular property prediction is an important process in designing and developing a new drug. Molecular property prediction is a process of predicting the property of a molecule from its molecular structure. Graph Neural Networks (GNNs) are a range of deep learning techniques that can be directly applied to graph data for prediction. Since the structure of the molecule is considered a graph, GNN is applied to predict the molecular property of the molecule. GNN includes two types, Message Passing Neural Networks (MPNN) and Graph Convolutional Networks (GCN). In this paper, the MPNN model and the GCN model are compared to determine the best model suitable for the prediction of BBB permeability. Based on the comparison of the GNN models, the MPNN model outperforms the GCN model with a prediction accuracyof 90%. [ABSTRACT FROM AUTHOR]

The term "spam" refers to any form of undesired and unsolicited electronic communication that is distributed in mass. Email is the most common medium for the transmission of spam, although it can also be transmitted via text message, phone call, or through social media. Spammers send large numbers of unwanted messages using a variety of different modes of communication. Some of them are marketing communications trying to sell you products that you have not requested. Other sorts of spam communications have the potential to infect your computer with malware, deceive you into disclosing personal information, or scare you into thinking that you need to pay to get out of a sticky situation. The demand for more dependable and efficient anti-spam filters has significantly increased as a result of the rise in unwanted email messages, known as spam. In other hand "Ham" refer to the desired and solicited electronic communication. In this project we made use of various reliable methods such as the TFIDF vectorizer and classifiers such as svm, naive-bayes, random forest, and XGBoost, all of which contributed to our efficiency rate of 96%. Comparative analysis of the benefits and drawbacks of currently available machine learning algorithms as well as unanswered research problems pertaining to spam filtering are presented in this paper. Deep learning and deep adversarial learning are two potential future tactics that we mentioned as ways to effectively combat spam email. [ABSTRACT FROM AUTHOR]

Search-based data analytics is a big area that requires great attention in many different areas. To address real-time issues, many methods and algorithms are utilised for search-based optimization. Many fields of study include design, accounting, finance, continuous imaging, and many more, but the most important are machine learning and deep learning. The experts are experimenting with a wide range of technologies and advancements, including open source, and have developed methods that enable them to achieve a greater degree of precision. When it comes to deep learning, it's inextricably linked to machine learning since it combines greater levels of execution and accuracy with a lower base error rate. Genetic Method, a high-performance algorithm for engineering optimization, is presented in the paper as use cases. For the purposes of data analytics and in particular for a multiprocessor scheduling strategy, the goal of this article is to provide the effective implementation of the famous evolutionary computing technique of genetic algorithm for genetic algorithms. Due to its better results on a variety of factors, the genetic algorithm described here may offer real and measurable results. There were no successful outcomes from the prior techniques' usage of greedy-based approaches, therefore researchers turned to genetic algorithm-based optimization. To design and simulate the implementation scenarios, we used MATLAB and Java-based development tools, and the results were very useful for optimizing. [ABSTRACT FROM AUTHOR]

Power plants become critical facilities that need to run every time without failure. The component in the power plant system has been equipped with sensors. These sensors will monitor the element's condition by periodically sending the related metric data to the control room or storing it for later evaluation. These sensor data can be analyzed for early anomaly detection on power plants. Traditional methods still need humans as experts to detect these anomalies; hence, human error is a factor. Using deep learning, we can take advantage of abundant data to learn the usual pattern and detect anomalies in future data. Many deep learning algorithms have been implemented in this case. However, the selection of these algorithms depends on the data's characteristics. This paper aims to provide a comprehensive, structured analysis of the data features and the algorithm used to detect anomalies in the data. Twenty-two recent power plant anomaly detection studies were compiled as a systematic review. This review will discuss power plant sensor data characteristics that need to be considered, such as raw data format, label usage, data dependence, and detection results. This review also discusses methods used to process unsupervised data, the relation of the multivariate data, and anomaly detection. This review aims to ease the selection of the appropriate anomaly detection algorithm for future research by considering the data characteristics. [ABSTRACT FROM AUTHOR]

In Today's internet world, detecting Fake News is facing a difficult task. Data integrity has been of maximum importance in the current news era due to the extensive range of free text. Therefore, generating and developing fraudulent information is more painless than ever. The spreading of such fake news influences society from diverse inappropriate views. Online Social Networks has become increasingly popular to obtain news and information on the internet due to their widespread uses. However, false information has been spread widely due to the internet's rapid growth. As a result, detecting fabricated information is critical in our present culture. As a resolution, in this paper, a Deep learning approach is suggested to confirm the trustworthiness of news. It is also proposed as a Multimodal faked information detection method that utilizes both the textual and graphical content of the data to determine its authenticity. The first step is to pretrain the BERT model to achieve the textual feature representation vector. As a second step, pretraining the VGG-19 model generates a visual characteristic representation vector. Two MCBP (Multimodal Compact Bilinear Pooling) modules have been created based on the proposed methodology. Graphic feature representation vectors with alerts are obtained using the first MCBP module. MCBP's second module combines the visual feature with the attention process and the vector of textual elements. With a second MCBP module, the attention process and a textual feature vector are utilized to input visual quality. You may be able to exploit the linked vector to your advantage in detecting fake information. In this article, a comparison is performed between the proposed method and two standard methods. EANN and Spot Fake approaches have been compared in this work, and it has been found that the proposed strategy in this article outperforms both in terms of accuracy, precision, and recall. [ABSTRACT FROM AUTHOR]

Polycystic Ovary Syndrome (PCOS) is a common hormonal disorder among women of reproductive age and is characterised by disrupted ovulation cycles and an increase in androgen secretion. It is allied with metabolic and cardiovascular risks as well as a greater risk of developing hormone dependent-cancer. Due to its high prevalence in the current society coupled with its potential hazards, its early and accurate diagnosis is of utmost importance. The diversity of symptoms often delays the process of confirmatory diagnosis and more than 50% of women remain undiagnosed. To address these issues by facilitating rapid detection, this article proposes early diagnosis of PCOS with hormonal test and ultrasound images using various Machine Learning and Deep Learning algorithms. In this paper, we have used Generative Adversarial Networks to extend our dataset, and Transfer Learning using Residual Networks (ResNets) and Inception Networks for the ultrasound images. For the hormonal dataset, after pre-processing, we have compared the performance of multiple models, where CatBoost produces the highest accuracy. [ABSTRACT FROM AUTHOR]

The global demand for energy is surpassing its availability owing to the increase in population and advancements in technology. It is essential to move closer to a dependable, economical, and sustainable renewable energy source to meet future energy demands. All alternative energy sources, including photovoltaic (PV) energy, have the potential to be renewable resources that can be used to address these ongoing issues with the energy crisis. PV systems are subject to the dynamic and fluctuating nature of the surrounding environment, which can lead to the manifestation of a diverse array of defects. It is essential to promptly identify and rectify any defects in order to ensure that the photovoltaic panel's efficiency, endurance, and durability remain uncompromised. The significance of automated defect detection is increasing in the PV industry due to the limitations of manual or visual inspection and the expanding production of PV modules. Therefore, this paper aims to develop a deep learning (DL) system that can accurately classify and detect defects in Electrouminescent (EL) images of PV cells, more specifically through implementing Convolutional Neural Networks CNN. Surprisingly, the suggested PV fault detection model outperformed commercially available PV fault detection systems by resolving existing challenges. The proposed model was further evaluated in terms of accuracy, loss, recall, f1 measure, and precision. The precision measure was 97.4% and the accuracy was 91.2%. [ABSTRACT FROM AUTHOR]

The convoluted and dynamic nature of cancer growth makes it difficult to predict long-term cancer survival. Evolutionary deep learning techniques have been growing in popularity as efficient and accurate prediction techniques in recent years. Long-term cancer survival prediction has been developed to utilize evolutionary deep-learning techniques. This research combines the power of evolutionary algorithms using Particle Swarm Optimization (PSO) with some deep learning techniques to improve the performance and accuracy of the deep learning models by optimizing its parameters and minimizing the error between the predicted and actual outputs. Overall, combining evolutionary algorithms with deep learning models for cancer survival analysis prediction demonstrates the potential to significantly improve the accuracy and efficiency of cancer survival predictions, leading to better treatment decisions and improved patient outcomes. This paper uses the gbsg2 survival analysis dataset, a well-known dataset commonly used in survival analysis studies. [ABSTRACT FROM AUTHOR]

False news is characterized as a made-up story with a goal to bamboozle or to delude. In this paper we present the answer for the errand of fake news discovery by using Deep Learning structures. Fake news identification is carried out utilizing two AI calculations, Natural Language Processing Algorithm(N=10) and Deep Learning algorithm(N=10) calculations. False and True these two sorts of dataset is utilized for Fake news recognition, and it is gathered from kaggle.com. Dataset comprises lines and 6 fundamental boundaries that are connected with the False news that information gathered from twitter. For each gathering more than 30 examples are taken, and it is separated into preparing and testing. Accuracy for Natural Language handling calculation is 91.300% and for Support Vector Machine calculation is 77.500%. There exists an insightful critical distinction between Natural Language Processing Technique and Support Vector Machine calculations with p<0.05. Fake news location utilizing Natural Language Processing calculation seems to acquire higher precision than the Support Vector Machine calculation. [ABSTRACT FROM AUTHOR]

Automated diagnosis of the cardiovascular disease from heartbeats of an individual is a difficult and challenging task. Essentially, it requires continuous monitoring system to track the irregular heartbeats in order to reduce the fatality rates. Electrocardiogram (ECG) is considered as one of the cheapest and most efficient clinical record used in hospitals to examine the cardiovascular condition of a heart patient. It is treated as a benchmark for sensing and identifying irregular heart rhythms. In modern era, due to abrupt increase in number of heart patients, cardiovascular health has become more important. In this paper, the aim of the research is to classify and predict the abnormal heartbeats from five different pre-defined classes such as, 'Fusion of Ventricular and Normal Beat', 'Supraventricular Premature Beat', 'Normal', 'Premature Ventricular Contraction' and 'Unclassifiable Beat'. The pre-trained VGG19 and DenseNet201 deep transferred learning models is used for classification and prediction. The proposed method is able to achieve overall 98% and 99% accuracy for VGG19 and DenseNet201 models, respectively. [ABSTRACT FROM AUTHOR]

Kidney is one of the significant organs in the human body which performs filtering out blood, balances fluid, removes the waste, maintains the level of electrolytes and hormone levels. So, any disorder or dysfunction in kidney needs to be detected on time in order to preserve life. Segmentation on kidney tumor in medical field is a critical task and many conventional methods have been employed for early prediction of kidney abnormalities but with limitations such as high cost, extended time for computation and analysis with huge amount of data. Due to all such problems, the prediction rate and accuracy has reduced considerably. In order to overcome the challenges, Artificial Intelligence (AI) technology has penetrated into the field of medicine particularly in the renal department. The evolution of AI in kidney therapies improve the process of diagnosis through several Machine Learning (ML) and Deep Learning (DL) algorithms. It has the capability of improving and influencing on the status with its capacity of learning from the massive data and apply them accordingly to differentiate on the circumstances. The storage of larger data and segmentation with AI assistance are highly helpful for the analysis of occurrence of the disease. AI algorithms have predicted the severity of tumor stages with effective accuracies. Hence, this paper provides a critical review of different AI based algorithms being used in the kidney tumor prognostication. Its numerous benefits in field of segmentation have been researched from the existing works and provides an insight on the contribution of AI in the kidney disease prediction. [ABSTRACT FROM AUTHOR]

Cashless transactions, which include internet, credit card, and mobile wallet transactions, are growing in popularity and have recently become well-known in financial transactions. The use of online transactions has increased, and there may be rapid improvement in e-trade and online banking in the coming decade. As the number of these cashless Transaction rises, so does the number of fraudulent transactions. Fraud can be identified by analyzing prior customer (user) spending patterns data on Transaction. This paper presents a review of various fraud detection techniques and discuss the issues regarding the financial dataset used in fraud detection technique. [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]

Physical disabilities cause significant distress in people's daily routine, especially for leg amputees as their movement is restricted. The civilized society has the responsibility to provide them with better prosthetics so they can live a normal life. Previously, prosthetic legs were developed for lower leg amputees. Today's electronic prosthesis provides a wide range of mobility compared to a mechanical prosthesis. Powered prosthetics are quite expensive and hence not widely used. We have proposed a low-cost control system above-knee powered prosthetics leg that is reliable, and the complexity of the model is less. The prostheses works by taking input from a sensor placed in the model and six FSR402 sensor and an IMU sensor that record force muscle activity of the thigh during variation of leg movement. The value of motion of the joint knee angle is determined by sensors. In this paper, we are discussing and comparing potential deep learning control system namely artificial neural network (ANN) as baseline and recurrent neural network-long short term memory (RNN-LSTM) to predict human gait cycle. A multi-layer perceptron (MLP) method is implemented. In general, network architecture consists of input layer which is represents by six inputs of the FSR402 sensor and an IMU sensor, fully connected layers, and an output layer as joint knee angle. As the result model of ANN as baseline has metric value MAE and R squared are 6.21 and 68.7% respectively whereas model of RNN-LSTM has metric value MAE and R squared are 4.99 and 96.9% respectively. The result indicates the RNN-LSTM model is most likely to perform fewer errors based on the dataset. RNN-LSTM outperforms the task with a greater value compared to the ANN model as the baseline. This present study suggests that deep learning with the recurrent neural network-long short term memory (RNN-LSTM) is suitable for the prediction of human gait cycle movement from the FSR402 and an IMU signal in the overall walking task. [ABSTRACT FROM AUTHOR]

This paper aims to develop a student worksheet with a discovery learning-STEM model on dynamic electric material to build on students' problem-solving skills in YAPPI Wonosari Vocational High school. The method used is R&D with a 4-D model that includes defining, designing, developing, and disseminating. The instruments used include validation sheets and tests—data in a physics worksheet to improve students' problem-solving skills. This study's data collection methods contain expert validation, observation, pre-test, and post-test—the data analysis using the sign test. The results showed that 1) the development of student worksheets with discovery learning-STEM on dynamic electric material was feasible in terms of the validation results. This can be seen from the results of validation that show the feasibility of physics worksheets for improving problem-solving skills is in the excellent category, and 2) the effectiveness of using physics worksheet with discovery learning-STEM based on data obtained from the pre-test and post-test results can improve problem-solving skills as a problem in the implementation class. Thus, student worksheet developed with discovery learning STEM is compatible with improving problem-solving skills. [ABSTRACT FROM AUTHOR]

The rapid developments of deep learning techniques are the process that can easily create and use a fake media known as Deepfakes. Deep learning techniques can generate facial motion transferring, gender-changing, and face swapping with others. Currently, it is possible to create hyper-realistic digital media like images and videos through deep learning techniques that are easily accessible. However, that becomes a threat for everyone, if used for harmful purposes. This work presents how deep learning is used for Deepfake generation and detection. There are two methods that are most popular for Deepfake generation: Generative Adversarial Networks (GANs) and autoencoder techniques that is presented in this work. It also, introduces a survey of some researchers' works and methods that used for Deepfake detection during the recent years. This paper proposed a method to discriminate if that video is real or fake by using a Support Vector Machine (SVM) classifier, the method presented in this work for deepfake detection is based on Gray Level Co-occurrence Matrix (GLCM) as a feature extraction method applying on 4 bands to obtain 13 feature form each one to achieve good accuracy of the classifier with a score 70.47% that batter than previous work that didn't exceed 60% of accuracy. [ABSTRACT FROM AUTHOR]

Coronavirus Disease (COVID-19) pandemic has emerged as one of the biggest challenge faced by researchers, especially in public health system. Since the virus spread and mutates quickly, strategies are needed so that the public health system doesn't collapse due to overcapacity. Modeling and forecasting accurately the COVID-19 daily new cases is very important to understand and help carry out risk management for the outbreak control. To obtain the estimation of the Indonesian daily cases of COVID-19 based on daily confirmed cases, we present a comparative study between three deep learning methods in this paper: Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), and Convolutional Long Short-Term Memory (Conv-LSTM). The data used is daily confirmed cases from September 1st to September 7th, 2021. The result is Conv-LSTM model produces better performance than other methods. Here we also provide 7 day out-of-sample forecast for daily cases from September 1st to September 7th, 2021 using each model. [ABSTRACT FROM AUTHOR]

Timely identification of diseases in crops and taking immediate actions to prevent them from being completely destroyed is a key part of agricultural production. Environmental challenges like unfavorable weather conditions and climate changes directly affect the plant growth and it causes viral infections that can lead to the eradication of the crops. Microorganisms such as bacteria and fungi can also cause various diseases in plants. In order to provide the accurate and better solution for a plant problem, the farmers first need to identify the exact kind of illness that the plant is having. The existing method of manual diagnosis of plant diseases by a field expert is time consuming as well as costly. Farmers cannot afford the cost of frequent visits to remote laboratories for diagnosis. The use of advanced and effective deep learning models can solve this problem and predict plant diseases at an early stage. IoT has become very popular in the agricultural industry as it provides state of the art solutions to plant problems. This paper aims to study the recent related works on this area and to analyze the advanced methods and models used in the implementation and to allow future research to learn larger capabilities of IoT and Deep Learning in proactively detecting the plant diseases with improved system performance and accuracy. [ABSTRACT FROM AUTHOR]

Techniques for detecting oral cancer are a developing area in medicine that includes cancer diagnosis and early detection. For the classification of oral cancer, we have tried to develop machine learning models like logistic regression (LR), naive Bayes (NB), K-Nearest Neighbors (KNN), support vector machine (SVM), decision tree (DT), and random forest (RF), as well as hybrid models like KNN-SVM and SVM-KNN and deep learning models like Xception, VGG, CNN, Inception, and MobileNet. A few researchers have already used the aforementioned models to categorize oral cancer using genetic data. Nevertheless, the database used for this study includes pictures of the lips and tongue that have been divided into malignant and non-cancerous categories. With the assistance of ENT doctors, the photos were categorized after being captured in various ENT hospitals in Ahmedabad. To determine which model is the most effective, the ML, hybrid, and DL models are evaluated based on numerous evaluation criteria. Experimental result shows that the accuracy, precision, recall, and Fl score of classification of the ML, Hybrid, and deep learning are discussed in this research paper. Out of which hybrid model SVM-KNN with PCA has 74%, Gradient boosting with PCA with 73%, and deep learning model VGG has I 00% accuracy. Hence, it may be concluded that Gradient boosting with PCA and advanced deep learning models like VGG, and Xception is the most favorable model for predicting the survival rate of oral cancer patients by classifying them into cancerous or non-cancerous. This research study motivates towards the early prediction comparatively reduces the severity, and economic cost as the treatment for early diagnosis is less. [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]

Soil is an essential part of the ornamental plant industry because the soil stores a lot of nutrients and microflora, which are very useful for plant growth. However, ordinary people or beginners in the ornamental plant industry sometimes do not understand the type of soil used, so they get less than optimal results in plant maintenance. Thus, this sometimes results in losses and decreased production income. So, to help the layman understand the type of soil that is good for use in the plant industry, this research will help classify the soil. In conducting soil classification, this research uses a method, namely CNN. CNN method is the most effective method in classifying. This paper aims to provide steps for the classification of 5 soil types. [ABSTRACT FROM AUTHOR]

In recent years, improvements in natural language processing and computer vision have come together to provide automatic image caption generation. Image captioning is the process of creating a description for an image. Captioning an image needs the recognition of significant items, their properties, and their connections within the image. Additionally, it must create phrases that are syntactically and semantically accurate. Deep learning approaches can address the complexities and difficulties associated with image captions. This paper describes a joint model which is capable of automatically captioning images using EfficientNet-B0 and a transformer with multi-head attention. The model is an aggregation of an EfficientNet & Transformer single encoder and decoder. The encoder utilizes EfficientNet-B0, a convolutional neural network-based algorithm that generates a detailed input image, represented by embedding them into a fixed-length vector. The decoder employs a transformer, and a multi-head attention mechanism to selectively concentrate attention on certain regions of images to predict the sentence. The proposed model was trained using a large dataset Flickr8k to optimize BLEU N-Gram (N=1,2,3,4), METEOR score, and CIDEr, to assess the probability of the target description phrase given in the training images. Our studies show that the proposed model can produce captions for images automatically. [ABSTRACT FROM AUTHOR]

The agricultural industry is the most prominent sector for the economic growth of the country. Cotton is an important crop that brings out major returns to the economy. In the current era, smart agriculture technology is incorporated which includes various modern techniques such as sensors, artificial intelligence, and computer vision to digitalize the whole process. This digitalization through modern technology enhances the fast and intelligent productivity of crops. In this paper convolutional neuralnetwork-based methods have been used to predict the maturity of the cotton balls using image features of mature and immature images. Real-time maturity detection is a challenging task for cotton fiber because of the many constraints in natural light conditionsand environmental effects. The proposed scheme uses annotation to classify the cotton balls. The classification is labeled as matureand immature images then the annotated images are used for training the deep learning models Resnet 50 and VGG 16. The averageaccuracy achieved from these methods is 92.47% for Resnet 50 and 96.58% for VGG16. The loss of the model is 21.05% for Resnet50 and 14.91% for VGG16. [ABSTRACT FROM AUTHOR]

The stock market serves as a vital investment platform, attracting numerous financial investors due to its growing capitalization. However, predicting stock prices accurately has always been a challenging task, requiring advanced algorithmic techniques. This research paper presents an innovative approach to improve stock market predictions by leveraging the strengths of Long Short-Term Memory (LSTM) based on a comparative analysis with Multi-Level Perceptron (MLP) models. The study compares the predictive capabilities of Random Forest, MLP, and LSTM models, with MLP demonstrating superior accuracy compared to Random Forest in the initial analysis. Building upon this finding, LSTM is employed by incorporating MLP predictions to forecast future market trends. The proposed technique demonstrates reliable results by combining machine learning and deep learning approaches. The MLP analysis further enhances the suggested model, resulting in improved accuracy and reduced mean square error. The result highlights the potential of integrating LSTM and MLP models for enhanced stock market predictions, offering valuable insights for financial forecasting and opening avenues for further research and refinement in this field. [ABSTRACT FROM AUTHOR]

This research paper explores the application of text classification and natural language processing techniques for enhancing hate speech detection. The study employs machine learning (ML) and deep learning models, including transformer models such as BERT, RoBERTa, and DistilBERT, to improve the accuracy of hate speech classifiers. Through a comprehensive empirical analysis on three diverse datasets (Data-ICWSM, Data-ALW2, Data-OLID), the study demonstrates the effectiveness of these models in accurately identifying hate speech. Compared to traditional baselines, the BERT models exhibit a significant performance boost in macro and weighted F1-scores. Additionally, the study addresses the challenge of imbalanced class distributions in the datasets by employing sampling techniques during training. Overall, the research highlights the potential of transformer models for hate speech detection and provides insights for future exploration in this domain. [ABSTRACT FROM AUTHOR]