Deep learning algorithms are rapidly changing the way in which audiovisual media can be produced. Synthetic audiovisual media generated with deep learning—often subsumed colloquially under the label “deepfakes”—have a number of impressive characteristics; they are increasingly trivial to produce, and can be indistinguishable from real sounds and images recorded with a sensor. Much attention has been dedicated to ethical concerns raised by this technological development. Here, I focus instead on a set of issues related to the (...) notion of synthetic audiovisual media, its place within a broader taxonomy of audiovisual media, and how deep learning techniques differ from more traditional approaches to media synthesis. After reviewing important etiological features of deep learning pipelines for media manipulation and generation, I argue that “deepfakes” and related synthetic media produced with such pipelines do not merely offer incremental improvements over previous methods, but challenge traditional taxonomical distinctions, and pave the way for genuinely novel kinds of audiovisual media. (shrink)

Recent work in mechanistic interpretability has studied how large language models recall facts stored in their weights. This paper argues that factual recall points to something broader: a general kind of operation in deep learning models, which I call feature recall. The core observation is that a linear projection can be read as retrieving stored information scaled by input activations. I define feature recall, show it applies across architectures, and contrast it with the established paradigm of feature combination. (...) I also consider how cases of feature recall might be mechanistically identified. The account offers philosophers a new conceptual tool for understanding deep learning and points to empirical research directions for mechanistic interpretability research. (shrink)

A corn crop typically refers to a large-scale cultivation of corn (also known as maize) for commercial purposes such as food production, animal feed, and industrial uses. Corn is one of the most widely grown crops in the world, and it is a major staple food for many cultures. Corn crops are grown in various regions of the world with different climates, soil types, and farming practices. In the United States, for example, the Midwest is known as the "Corn Belt" (...) due to its extensive corn production. Corn crops can be grown using a variety of methods, including conventional tillage, no-till farming, and the use of genetically modified crops. Due to its importance, there is a need for discovering the corn diseases and treating them. The aim of this study is to propose a deep learning model for the classification of corn leaf diseases based on Convolutional Neural Network. A model for classifying maize diseases was developed using a dataset for classifying Corn diseases that contains 4 classes of disease. The dataset was collected from Keggel website. The proposed model was trained, validated, and tested. The F1-score (99.83%), Recall (99.83%), precision (99.83%), Accuracy (99.83%). (shrink)

We claim that scientists working with deep learning (DL) models exhibit a form of pragmatic understanding that is not reducible to or dependent on explanation. This pragmatic understanding comprises a set of learned methodological principles that underlie DL model design-choices and secure their reliability. We illustrate this action-oriented pragmatic understanding with a case study of AlphaFold2, highlighting the interplay between background knowledge of a problem and methodological choices involving techniques for constraining how a model learns from data. Building (...) successful models requires pragmatic understanding to apply modelling strategies that encourage the model to learn data patterns that will facilitate reliable generalisation. (shrink)

The COVID-19 outbreak has had a significant influence on the health of people all across the world, and preventing its further spread requires an early and correct diagnosis. Imaging using X-rays is often used to identify respiratory disorders like COVID-19, and approaches based on machine learning may be used to automate the diagnostic process. In this research, we present a deep learning approach for COVID-19 identification in X-ray pictures utilizing global thresholding. Our framework consists of two main (...) components: (1) global thresholding to preprocess X-ray images and extract features, and (2) a Convolutional Neural Network (CNN) for classifying the images as positive (indicating COVID-19) or negative (indicating no COVID-19). Global thresholding is used to convert X-ray images into binary images, which can highlight the features of interest, such as lung opacity. The Network is instructed to use a dataset consisting of X-ray pictures that have been labeled with the COVID-19 status of each image. The usefulness of our approach was shown by experimental findings obtained using a dataset of X-ray pictures that were made accessible to the public. We achieve high accuracy in COVID-19 detection, outperforming traditional machine learning techniques. Our framework is also able to correctly identify COVID-19 in images that were misclassified by human experts. (shrink)

The ever-evolving ways attacker continues to improve their phishing techniques to bypass existing state-of-the-art phishing detection methods pose a mountain of challenges to researchers in both industry and academia research due to the inability of current approaches to detect complex phishing attack. Thus, current anti-phishing methods remain vulnerable to complex phishing because of the increasingly sophistication tactics adopted by attacker coupled with the rate at which new tactics are being developed to evade detection. In this research, we proposed an adaptable (...) framework that combines Deep learning and Randon Forest to read images, synthesize speech from deep-fake videos, and natural language processing at various predictions layered to significantly increase the performance of machine learning models for phishing attack detection. (shrink)

Rainfall prediction is one of the challenging tasks in weather forecasting. Accurate and timely rainfall prediction can be very helpful to take effective security measures in dvance regarding: on-going construction projects, transportation activities, agricultural tasks, flight operations and flood situation, etc. Data mining techniques can effectively predict the rainfall by extracting the hidden patterns among available features of past weather data. This research contributes by providing a critical analysis and review of latest data mining techniques, used for rainfall prediction. In (...) our proposed system we propose a new forecasting method that uses a Convolutional Neural Network monthly rainfall for a selected location. In our proposed system we are going to forecast the rainfall result based on the mean square error, mean absolute error and root mean square error, which we get in train and test of the dataset based deep learning technique. (shrink)

: Background: Vegetable agriculture is very important to human continued existence and remains a key driver of many economies worldwide, especially in underdeveloped and developing economies. Objectives: There is an increasing demand for food and cash crops, due to the increasing in world population and the challenges enforced by climate modifications, there is an urgent need to increase plant production while reducing costs. Methods: In this paper, Lemon classification approach is presented with a dataset that contains approximately 2,000 images belong (...) to 3 species at a few developing phases. Convolutional Neural Network (CNN) algorithms, a deep learning technique extensively applied to image recognition was used, for this task. The results: found that CNN-driven lemon classification applications when used in farming automation have the latent to enhance crop harvest and improve output and productivity when designed properly. The trained model achieved an accuracy of 99.48% on a held-out test set, demonstrating the feasibility of this approach. (shrink)

Real-time monitoring systems use surveillance videos to automatically detect motorcycle helmet requirements and triple-riding violations, which protect road safety. Deep learning methods currently show practical worth for addressing surveillance system constraints because they have developed superior capabilities in object detection and classification. The models deliver poor results repeatedly because they are limited by low-resolution video, together with adverse weather conditions, as well as problems from occlusions and deficient illumination conditions. The issue of recognizing multiple individuals riding together on (...) one motorcycle requires better solution methods because it remains unresolved. YOLOv8 operates in real-time to detect both helmet violations and multiple riders on vehicles using CCTV video and webcam feeds. The detection system reaches a 95% accuracy level, thereby establishing itself as an effective monitoring method in automated traffic control. A YOLOv8 detection model serves for detecting triple riders, along with YOLOv8-based detection of helmet violations. The model has tailored detection skills since its training with helmet-specific data enables it to monitor properly non-helmeted riders. The system supplies capabilities for processing video and image data, which helps it create visual violation marks for practical realworld applications. Quality evaluation of the system happens through precision and recall metrics, which measure its reliable operation functions. The method utilizes publicly available data and self-collected information to solve illumination conditions and issues that affect detection systems. The evolution of traffic rule enforcement systems has improved thanks to the resolution of implementation problems. The research application proves how deep learning transforms surveillance systems, which advances road safety through system improvements. Better traffic management possibilities, combined with improved efficiency and innovation, enable wider expansion of safety programs in urban areas. (shrink)

Neural networks and deep learning have significantly advanced the field of artificial intelligence, offering solutions to complex problems such as image recognition, natural language processing, and decision-making tasks. This paper explores the principles of neural networks, particularly deep learning models, their evolution, applications, challenges, and potential future directions. A comprehensive analysis of key algorithms, architectures, and advancements is provided, with an emphasis on the practical implications of deep learning in various domains. By understanding the (...) foundations of neural networks, we can better address issues of scalability, interpretability, and computational efficiency. (shrink)

Artificial intelligence (AI) is an area of computer science that emphasizes the creation of intelligent machines or software that work and react like humans. Some of the activities computers with artificial intelligence are designed for include, Speech, recognition, Learning, Planning and Problem solving. Deep learning is a collection of algorithms used in machine learning, It is part of a broad family of methods used for machine learning that are based on learning representations of data. (...) Deep learning is a technique used to produce Pneumonia detection and classification models using x-ray imaging for rapid and easy detection and identification of pneumonia. In this thesis, we review ways and mechanisms to use deep learning techniques to produce a model for Pneumonia detection. The goal is find a good and effective way to detect pneumonia based on X-rays to help the chest doctor in decision-making easily and accuracy and speed. The model will be designed and implemented, including both Dataset of image and Pneumonia detection through the use of Deep learning algorithms based on neural networks. The test and evaluation will be applied to a range of chest x-ray images and the results will be presented in detail and discussed. This thesis uses deep learning to detect pneumonia and its classification. (shrink)

There are many eye diseases but the most two common retinal diseases are Age-Related Macular Degeneration (AMD), which the sharp, central vision and a leading cause of vision loss among people age 50 and older, there are two types of AMD are wet AMD and DRUSEN. Diabetic Macular Edema (DME), which is a complication of diabetes caused by fluid accumulation in the macula that can affect the fovea. If it is left untreated it may cause vision loss. Therefore, early detection (...) of diseases is a critical importance. Our main goal is to help doctors detect these diseases quickly before reaching a late stage of the disease. In ophthalmology, optical coherence tomography (OCT) is critical for evaluating retinal conditions. OCT is an imaging technique used to capture high-resolution cross-sections of the retinas of patient. In this thesis, we review ways and techniques to use deep learning classification of the optical coherence tomography images of diseases from which a Retinal is suffering. The models used to improve patient care are (VGG-16, MobileNet, ResNet-50, Inception V3, and Xception) to reduce costs and allow fast and reliable analysis in large studies. The obtained results are encouraging, since the best model ResNet-50 reaching 96.21% of testing accuracy, which is very useful for doctors, to diagnose retinal diseases. (shrink)

This article addresses computational procedures that are no longer constrained by human modes of representation and considers how these procedures could be philosophically understood in terms of ‘algorithmic thought’. Research in deep learning is its case study. This artificial intelligence (AI) technique operates in computational ways that are often opaque. Such a black-box character demands rethinking the abstractive operations of deep learning. The article does so by entering debates about explainability in AI and assessing how technoscience (...) and technoculture tackle the possibility to ‘re-present’ the algorithmic procedures of feature extraction and feature learning to the human mind. The article thus mobilises the notion of incommensurability (originally developed in the philosophy of science) to address explainability as a communicational and representational issue, which challenges phenomenological and existential modes of comparison between human and algorithmic ‘thinking’ operations. (shrink)

The global cultivation of grapes reaches approximately 77.8 million tons annually, according to the International Organization of Vine and Wine. While grapes remain a vital agricultural commodity and dietary staple worldwide, their production faces serious threats from common diseases like black rot, Esca, and leaf blight. Current disease detection methods in modern vineyards primarily depend on manual visual inspection, a practice that often delays diagnosis and leads to reduced yields and compromised fruit quality. The integration of automated detection methods, particularly (...) those based on machine learning, is crucial for promoting sustainable viticulture. This study leverages the power of deep learning, a subfield of machine learning particularly adept at interpreting image data. Specifically, we implemented Convolutional Neural Networks (CNNs) to classify images of grapevine leaves as either healthy or diseased. The models employed included a custom-built baseline CNN and a range of advanced transfer learning models: DenseNet121, EfficientNetB7, MobileNetV2, ResNet50, and VGG16. While we initially hypothesized that ResNet50 would yield the highest accuracy due to its deep architecture, experimental results revealed that EfficientNetB7 outperformed all other models. To further enhance classification performance, we constructed a max-voting ensemble using the top three performing models. This ensemble approach outmatched the performance of individual models. The final model was deployed via a web-based interface, enabling vineyard professionals and growers to detect black rot, Esca, leaf blight, or confirm leaf health by uploading images captured in real-world vineyard environments. (shrink)

The infectious illness known as Pneumonia is regularly a result of contamination because of a bacterium in the alveoli of the lungs. While an infected tissue of the lungs has an infection, it builds up pus in it. To find out if the patient has those illnesses, professionals perform bodily exams and diagnose their patients through Chest X-ray, ultrasound, or biopsy of lungs. Misdiagnosis, erroneous treatment, and if the disease is overlooked will result in the patient’s lack of lifestyle. The (...) progression of Deep learning contributes to aid in the decision-making procedure of specialists to diagnose sufferers with these illnesses. The look employs a bendy and efficient technique of deep learning of applying the model of CNN in predicting and detecting a patient unaffected and affected with the sickness using a chest X-ray photograph. The take a look at utilizing an accrued dataset of 20,000 images using a 256x256 photograph decision with 32 batch length is applied to prove the overall performance of the CNN model being educated. The trained-version produced an accuracy charge of 95% at some point of the overall performance training. Based on the end result of the experiment carried out, the research study can detect and are expecting COVID-19, bacterial, and viral-pneumonia sicknesses based totally on chest X-ray photos. (shrink)

Crop diseases pose several challenges in the agricultural industry. Plant diseases can have a devastating impact on both yield and quality loss. This project presents a deep-learning based sugarcane disease classification and alert system to facilitate machine detection of disease, as well as actions to take in response to diagnosis. A dataset of images of sugarcane leaves, was modified through advanced pre-processing techniques such as cropping, rotating, image enhancement, detections edges, and adjusting for wavy images. The techniques used (...) for data augmentation helped address class imbalance as well as overfitting. The classification model implemented for this project is the EfficientNet architectures to produce better feature extraction and predictions. The classification model was trained on the pre-processed and augmented dataset and reached 97 percent accuracy on the test data after 150 epochs. The explainability of the classifying model was provided through LIME (Local Interpretable Model-Agnostic Explanations), which allowed users to see the explanation for classification for an individual image. The smart classifying system was developed as a web application (UI) and allows users to input their image to be diagnosed. Once diagnosed, the model classifies the disease and detail actions, including: recommended treatments and precautionary actions. This system shows the power of artificial intelligence to enhance management of diseases in agriculture, while providing a scalable user-friendly environment to help faciliate on farm decision-making regarding crop health and yield. (shrink)

Computational complexity is a discipline of computer science and mathematics which classifies computational problems depending on their inherent difficulty, i.e. categorizes algorithms according to their performance, and relates these classes to each other. P problems are a class of computational problems that can be solved in polynomial time using a deterministic Turing machine while solutions to NP problems can be verified in polynomial time, but we still do not know whether they can be solved in polynomial time as well. A (...) solution for the so-called NP-complete problems will also be a solution for any other such problems. Its artificial-intelligence analogue is the class of AI-complete problems, for which a complete mathematical formalization still does not exist. In this chapter we will focus on analysing computational classes to better understand possible formalizations of AI-complete problems, and to see whether a universal algorithm, such as a Turing test, could exist for all AI-complete problems. In order to better observe how modern computer science tries to deal with computational complexity issues, we present several different deep-learning strategies involving optimization methods to see that the inability to exactly solve a problem from a higher order computational class does not mean there is not a satisfactory solution using state-of-the-art machine-learning techniques. Such methods are compared to philosophical issues and psychological research regarding human abilities of solving analogous NP-complete problems, to fortify the claim that we do not need to have an exact and correct way of solving AI-complete problems to nevertheless possibly achieve the notion of strong AI. (shrink)

Tomatoes are part of the major crops in food security. Tomatoes are plants grown in temperate and hot regions of South American origin from Peru, and then spread to most countries of the world. Tomatoes contain a lot of vitamin C and mineral salts, and are recommended for people with constipation, diabetes and patients with heart and body diseases. Studies and scientific studies have proven the importance of eating tomato juice in reducing the activity of platelets in diabetics, which helps (...) in protecting them from developing deadly blood clots. A tomato classification approach is presented with a data set containing approximately 5,266 images with 7 species belonging to tomatoes. The Neural Network Algorithms (CNN), a deep learning technique applied widely in image recognition, is used for this task. (shrink)

The proposed system leverages Convolutional Neural Network (CNN) technologies to perform early lung cancer detection through the analysis of chest CT scan images. Designed as an AI-powered diagnostic aid, the system operates independently of human interpretation, enhancing speed and reliability in medical imaging evaluation. Built using Python-based technologies, the system utilizes TensorFlow and Keras frameworks for deep learning model development and training, supported by OpenCV for image pre-processing and visualization. The system processes CT scan slices to identify pulmonary (...) nodules and classifies them into benign or malignant categories using a trained CNN model. The detection pipeline includes grayscale normalization, image resizing, and segmentation procedures to enhance the visibility of lung tissues. The CNN model extracts spatial hierarchies of features through multiple convolutional and pooling layers, followed by fully connected layers for final classification. In cases where the system detects potential malignancy, it generates a detailed report including confidence scores, prediction labels, and scan timestamps. To maintain responsiveness and scalability, prediction requests are managed through concurrent processing threads, ensuring timely feedback even when multiple scans are evaluated in sequence. An alert throttling mechanism ensures the reduction of false positives by incorporating prediction stability checks over multiple frames or scans. (shrink)

Crop diseases pose several challenges in the agricultural industry. Plant diseases can have a devastating impact on both yield and quality loss. This project presents a deep-learning based sugarcane disease classification and alert system to facilitate machine detection of disease, as well as actions to take in response to diagnosis. A dataset of images of sugarcane leaves, was modified through advanced pre-processing techniques such as cropping, rotating, image enhancement, detections edges, and adjusting for wavy images. The techniques used (...) for data augmentation helped address class imbalance as well as overfitting. The classification model implemented for this project is the EfficientNet architectures to produce better feature extraction and predictions. The classification model was trained on the pre-processed and augmented dataset and reached 97 percent accuracy on the test data after 150 epochs. The explainability of the classifying model was provided through LIME (Local Interpretable Model-Agnostic Explanations), which allowed users to see the explanation for classification for an individual image. The smart classifying system was developed as a web application (UI) and allows users to input their image to be diagnosed. Once diagnosed, the model classifies the disease and detail actions, including: recommended treatments and precautionary actions. This system shows the power of artificial intelligence to enhance management of diseases in agriculture, while providing a scalable user-friendly environment to help faciliate on farm decision-making regarding crop health and yield. (shrink)

Abstract cantaloupe and honeydew melons are part of the muskmelon family, which originated in the Middle East. When picking either cantaloupe or honeydew melons to eat, you should choose a firm fruit that is heavy for its size, with no obvious signs of bruising. They can be stored at room temperature until you cut them, after which they should be kept in the refrigerator in an airtight container for up to five days. You should always wash and scrub the rind (...) of your melon before you cut it to remove any dirt or bacteria on the outside. In this paper, cantaloupe classification approach is presented with a dataset that contains approximately 1,312 of Cantaloupes and honeydews. Convolutional Neural Network (CNN) algorithms, a deep learning technique extensively applied to image recognition was used, for this task. The results found that CNN-driven classification applications when used in farming automation have the latent to enhance crop harvest and improve output and productivity when designed properly. The trained model achieved an accuracy of 99.74% on a held-out test set, demonstrating the feasibility of this approach. (shrink)

Artificial intelligence (AI), deep learning, machine learning and neural networks represent extremely exciting and powerful machine learning-based techniques used to solve many real-world problems. Artificial intelligence is the branch of computer sciences that emphasizes the development of intelligent machines, thinking and working like humans. For example, recognition, problem-solving, learning, visual perception, decision-making and planning. Deep learning is a subset of machine learning in artificial intelligence that has networks capable of learning unsupervised (...) from data that is unstructured or unlabeled. Deep learning is a technique used to generate face detection and recognize it for real or fake by using profile images and determine the differences between them. In this study, we used deep learning techniques to generate models for Real and Fake face detection. The goal is determining a suitable way to detect real and fake faces. The model was designed and implemented, including both Dataset of images: Real and Fake faces detection through the use of Deep learning algorithms based on neural networks. We have trained dataset which consists of 9,000 images for total in 150 epochs, and got the ResNet50 model to be the best model of network architectures used with 100% training accuracy, 99.18% validation accuracy, training loss 0.0003, validation loss 0.0265, and testing accuracy 99%. (shrink)

Distributional Semantic Models (DSM) that represent words as vectors of weights over a high dimensional feature space have proved very effective in representing semantic or syntactic word similarity. For certain tasks however it is important to represent contrasting aspects such as polarity, opposite senses or idiomatic use of words. We present a method for computing discriminative word embeddings can be used in sentiment classification or any other task where one needs to discriminate between contrasting semantic aspects. We present an experiment (...) in the identification of reports on natural disasters in tweets by means of these embeddings. (shrink)

Abstract: Potatoes are edible tubers, available worldwide and all year long. They are relatively cheap to grow, rich in nutrients, and they can make a delicious treat. The humble potato has fallen in popularity in recent years, due to the interest in low-carb foods. However, the fiber, vitamins, minerals, and phytochemicals it provides can help ward off disease and benefit human health. They are an important staple food in many countries around the world. There are an estimated 200 varieties of (...) potatoes, which can be classified into a number of categories based on the cooked texture and ingredient functionality. Using a public dataset of 2400 images of potatoes, we trained a deep convolutional neural network to identify 4 types (Red, Red Washed, Sweet, and White).The trained model achieved an accuracy of 99.5% of test set, demonstrating the feasibility of this approach. (shrink)

Suspicious Activity Detection System is a real-time monitoring and analysis tool developed using Python with Tkinter, OpenCV, and Image AI libraries. The system integrates a modern graphical user interface (GUI) to facilitate video processing, frame generation, and suspicious activity detection from both live camera feeds and prerecorded CCTV footage. Leveraging the DenseNet121 deep learning model for image classification, the system processes video frames to identify potential suspicious activities, employing temporal consistency and confidence thresholding to enhance detection accuracy. Key (...) features include live feed toggling, video recording with timestamped storage, and automated frame extraction, with suspicious frames flagged and saved for further review. The GUI, styled with a sleek dark theme and gradient aesthetics, provides intuitive controls and dual text areas for logging frame generation and detection results. Advanced visualization overlays detection metrics on frames, while the system ensures robust error handling and resource cleanup. Designed for security applications, this tool demonstrates a scalable approach to real-time video analysis, balancing usability and performance. Future enhancements could include multicamera support and expanded activity classification capabilities. (shrink)

The deep learning methods for dealing with multi-class imbalance in water quality data are comprehensively examined in this paper. The findings show a notable increase in publications, especially since 2021, and a clear preference for deep learning techniques when classifying imbalanced data. Background: A thorough review of the body of literature included articles from significant digital libraries that were published between January 2012 and December 2024. Based on several important factors, such as commonly used datasets and (...) types, years of publication, different sources, research and empirical types, assessment measures, and development tools, 59 articles were chosen and examined. Objective: This study examines how deep learning techniques respond to data that is unbalanced when there are multiple classes and how the deep learning models' large capacity and intricate structures make them appropriate for these kinds of tasks, with an emphasis on water datasets. Conclusion: The paper emphasizes the difficulties with data diversity and computing efficiency, along with possible solutions for reliable real-world applications. It also discusses innovative solutions that enhance the reliability of real-world applications. (shrink)

Text classification is one of the most important task in natural language processing, In this research, we carried out several experimental research on three (3) of the most popular Text classification NLP classifier in Convolutional Neural Network (CNN), Multinomial Naive Bayes (MNB), and Support Vector Machine (SVN). In the presence of enough training data, Deep Learning CNN work best in all parameters for evaluation with 77% accuracy, followed by SVM with accuracy of 76%, and multinomial Bayes with least (...) performance of 69% accuracy. CNN has the best performance in the presence of large enough training dataset because of the presence of filter/ kernels which help to indentify patterns in text data regardless of their position in the sentence. We repeated the training again with just one-third of our data, at this point SVM comes with the best performance, the performance of CNN noticeably drops but still better than Multinomial Naive Bayes, the reason why SVM works best when we reduce the training data was because of its ability to look for a hyper-plane that creates a boundary between different classes of data so as to properly classify them, so we believed that getting the hyper-plane was more efficient when we reduce the dataset, hence reason for the good performance. Multinomial Naive Bayes have the least performance which we attributed to its assumption of independency between the features which sometimes does not hold true. We concluded that availability of data should be an important factor when choosing classifier for Natural Language Processing Text Classification task. CNN should be use in the presence of enough dataset, and SVM should be use when data is not enough. Multinomial Naive Bayes must not be trusted with state of the art NLP task due to its assumption of independency between the features. (shrink)

By Agriculture plays a major role in developing countries like India, however the food security still remains a vital issue. Most of the crops get wasted due to lack of storage facility, transportation, and plant diseases. More than 15% of the crops get wasted in India due to diseases and hence it has become one of the major concern to be resolved. There is a need of automatic system that can identify these diseases and help farmers to take appropriate steps (...) to get rid of crop loss. Farmers have followed the conventional method of identifying the plant disease with their naked eyes, and it not possible for all the farmers to identify these diseases the same way. With the advance in Artificial Intelligence, there is a need to incorporate the facilities of the computer vision in the field of agriculture. Deep Learning rich libraries and user as well as developer friendly environment to work with, all these qualities make Deep Learning as the favorable method to get started with this problem. In this paper we have used Deep Learning because of the advantages it offers to work with images especially in image classification to get improvised results. The methodology includes taking leaves of infected crops and label them as per the disease pattern. The images of infected leaves are processed pixel based operations are applied to improve the information from the image. As a next step feature extraction is done followed by image segmentation and at the last classification of crop diseases based on the patterns extracted from the diseased leaves. The CNN (Convolutional Neural Network) is used for the classification of diseases, for the demonstration purpose the public dataset is used consisting of around 87 K images (RGB type images) including healthy as well as diseased leaves. (shrink)

Near the end of December 2019, the globe was hit with a major crisis, which is nothing but the coronavirusbased pandemic. The authorities at the train station should also keep in mind the need to limit the spread of the covid virus in the event of a global pandemic. When it comes to controlling the COVID-19 epidemic, public transportation facilities like train stations play a pivotal role because of the proximity of so many people who may be exposed to the (...) virus. Using common place CCTV cameras and deep learning with simple online and real-time (DeepSORT) methods, this study develops social distance monitoring using a YOLOv4 identification of a Surveillance Object Model. Based on experiments conducted with a minicomputer equipped with an Intel 11th Gen Intel(R) Core(TM) i3-1115G4 at 3.00GHz, 2995 Mhz, two Core(s), four Logical processor, four gigabytes of random-access memory (RAM), this paper makes use of CCTV surveillance, which was put into practice at the Guindy railway station, Chennai, Tamilnadu in India in order to detect the violation of social distancing. (shrink)

Customer Relationship Management has become one of the most valuable tools in business, with the help of predictive analyzes organizations can identify their customer’s needs and improve business results. This research focuses on using of predictive analytics in Salesforce, the prominent CRM system that helps to gain insights from large volumes of data with the help of deep learning approaches. Using current advanced models like Recurrent Neural Networks and the transformer connections, companies can gain deep insights into (...) such patterns and trends as customer buying behavior, sales anticipation, and customer lifetime value. It is not only effective in improving decision making but also useful in the customization of the marketing strategy involving the delivery of messages that the customer might be interested. Integrating deep learning into Salesforce’s analytics is a revolution in the CRM system function. Former approaches in data analysis could not deal with unstructured data or analyses customer paths, as mentioned before. However, thanks to deep learning, large-scale customer data and aspects of previously hidden patterns have become changer for business interpretation. This research demonstrates the potential of using deep learning for predictive analytics to inform customer-oriented initiatives and increase the proportion of customers amongst leads and stability of client base. Thus, by focusing on Salesforce as an example, this study gives a clear vision of how artificial intelligence revolutionizes the approach to CRM, serves as the reference point for data-driven businesses in the digital environment. (shrink)

The banking sector faces growing challenges in identifying and managing risks due to the complexity of financial transactions and increasing fraud. This research presents a framework that combines multiple agents with deep learning to improve risk prediction in banking. Each agent focuses on specific tasks like cleaning data, selecting important features, and detecting unusual activities, ensuring a detailed risk assessment. A deep learning model is used to analyze large amounts of transaction data and identify patterns that (...) may signal potential risks. Tests on real-world banking data show that this approach is more accurate, faster, and effective than traditional methods. By combining the strengths of multi-agent systems and deep learning, this study offers a reliable and flexible solution to help banks manage and adapt to changing risks more effectively. (shrink)

Deep learning has emerged as a transformative technology, enabling advancements in fields such as computer vision, natural language processing, and autonomous systems. Python, with its comprehensive libraries and frameworks, has become the primary language for developing deep learning models. This paper explores the latest advancements in Python-based deep learning, focusing on key frameworks, algorithms, and innovations. It also discusses the challenges associated with implementing deep learning solutions, such as computational cost, data quality, (...) and model interpretability. Furthermore, it addresses sustainable approaches to deep learning, emphasizing energy-efficient techniques, model optimization, and the adoption of green computing practices. By understanding these advancements and challenges, we can push forward towards more efficient and sustainable deep learning solutions using Python. (shrink)

Given the growing dependence on digital systems and the escalation of financial fraud occurrences, it is imperative to implement efficient cyber security protocols and fraud detection methodologies. The threat's dynamic nature often challenges conventional methods, necessitating the adoption of more sophisticated strategies. Individuals depend on pre-established regulations or problem-solving processes, which possess constraints in identifying novel and intricate fraudulent trends. Conventional techniques need help handling noise data and the substantial expenses incurred by false positives and true positives. To tackle these (...) obstacles, the present study introduces Deep Fraud Net, a framework that utilizes deep learning to detect and classify instances of financial fraud and cyber security threats. The Deep Fraud Net system model entails the utilization of a deep neural network to acquire intricate patterns and characteristics from extensive datasets through training. The framework integrates noise reduction methods to enhance the precision of fraud detection and improve the quality of input data. The Deep Fraud Net method attains a precision of 98.85%, accuracy of 93.35%, sensitivity of 99.05%, specificity of 93.16%, false positive rate of 7.34%, and true positive rate of 89.58%. The findings suggest that Deep Fraud Net can effectively detect and categorize instances of fraudulent behavior with a reduced occurrence of misclassifications. The method exhibits potential implications for diverse domains that prioritize robust security and fraud detection, including but not limited to banking, e-commerce, and online transactions. (shrink)

Every person has his/her own unique signature that is used mainly for the purposes of personal identification and verification of important documents or legal transactions. There are two kinds of signature verification: static and dynamic. Static(off-line) verification is the process of verifying an electronic or document signature after it has been made, while dynamic(on-line) verification takes place as a person creates his/her signature on a digital tablet or a similar device. Offline signature verification is not efficient and slow for a (...) large number of documents. To overcome the drawbacks of offline signature verification, we have seen a growth in online biometric personal verification such as fingerprints, eye scan etc. In this paper we created CNN model using python for offline signature and after training and validating, the accuracy of testing was 99.70%. (shrink)

Computer vision, or image recognition, analyses and interprets visual data in real-world scenarios like images and videos. AI and ML research focusses on object, scene, action, and feature identification because of its usefulness in image processing. Neural networks and deep learning have improved image recognition systems significantly in recent years. Early image recognition used template matching to identify objects. A photo is compared to a stored template using similarity measures like correlation to get the best match. There are (...) several constraints, especially with distorted, scaled, or noisy images. Template matching is computationally expensive. CNNs are the most common deep learning architecture for computer vision. CNNs automatically learn visual input hierarchies. They include pooling, convolutional, and fully connected layers. Regional patterns like shapes, textures, and edges in the input image are filtered by convolutional layers. Pooling layers reduce feature map spatial dimensions, making feature extraction consistent and robust. Thick layers, or fully linked layers, classify or regress using learning features. CNN training requires a big labelled dataset. This research article introduces a deep learning model for the classification and recognition of images. The input data for this model is the Images Data Set. The GenNet Algorithm is employed to derive critical features. The accuracy of classification results is enhanced through the use of preprocessing and feature extraction. The classification model is generated using Convolution Neural Network, AlexNet, ResNet, and VGG. (shrink)

Fruits are a rich source of energy, minerals and vitamins. They also contain fiber. There are many fruits types such as: Apple and pears, Citrus, Stone fruit, Tropical and exotic, Berries, Melons, Tomatoes and avocado. Classification of fruits can be used in many applications, whether industrial or in agriculture or services, for example, it can help the cashier in the hyper mall to determine the price and type of fruit and also may help some people to determining whether a certain (...) type of fruit meets their nutritional requirement. In this paper, machine learning based approach is presented for classifying and identifying 10 different fruit with a dataset that contains 6847 images use 4793 images for training, 1027 images for validation and 1027 images for testing. A deep learning technique that extensively applied to image recognition was used. We used 70% from image for training and 15% from image for validation 15% for testing. Our trained model achieved an accuracy of 100% on a held-out test set, demonstrating the feasibility of this approach. (shrink)

A new trend in deep learning, represented by Mutual Information Neural Estimation (MINE) and Information Noise Contrast Estimation (InfoNCE), is emerging. In this trend, similarity functions and Estimated Mutual Information (EMI) are used as learning and objective functions. Coincidentally, EMI is essentially the same as Semantic Mutual Information (SeMI) proposed by the author 30 years ago. This paper first reviews the evolutionary histories of semantic information measures and learning functions. Then, it briefly introduces the author’s semantic (...) information G theory with the rate-fidelity function R(G) (G denotes SeMI, and R(G) extends R(D)) and its applications to multi-label learning, the maximum Mutual Information (MI) classification, and mixture models. Then it discusses how we should understand the relationship between SeMI and Shannon’s MI, two generalized entropies (fuzzy entropy and coverage entropy), Autoencoders, Gibbs distributions, and partition functions from the perspective of the R(G) function or the G theory. An important conclusion is that mixture models and Restricted Boltzmann Machines converge because SeMI is maximized, and Shannon’s MI is minimized, making information efficiency G/R close to 1. A potential opportunity is to simplify deep learning by using Gaussian channel mixture models for pre-training deep neural networks’ latent layers without considering gradients. It also discusses how the SeMI measure is used as the reward function (reflecting purposiveness) for reinforcement learning. The G theory helps interpret deep learning but is far from enough. Combining semantic information theory and deep learning will accelerate their development. (shrink)

In computer science, Artificial Intelligence (AI), sometimes called machine intelligence, is intelligence demonstrated by machines, in contrast to the natural intelligence displayed by humans and other animals. Computer science defines AI research as the study of "intelligent agents": any device that perceives its environment and takes actions that maximize its chance of successfully achieving its goals. Deep Learning is a new field of research. One of the branches of Artificial Intelligence Science deals with the creation of theories and (...) algorithms that allow the machine to learn by simulating neurons in the human body. Most in-depth learning research focuses on finding high-level methods. The strippers analyze a large data set using linear and nonlinear transformations. The method of deep learning is used in the detection of several diseases including blood cell diseases and their classification using the radiography of blood cells to help decision makers to know the type of blood cell and its associated diseases and the results will be presented in detail and discussed. This thesis is using python language and deep learning to detect blood cell diseases and their classifications. The proposed deep learning model was trained, validated and the tested. The accuracy of proposed model was 98.00%. (shrink)

Contemporary philosophy of artificial intelligence is frequently paralysed by binary debates regarding machine sentience ("the hard problem of consciousness") and the regulatory demand for algorithmic explainability ("opening the black box"). This article posits that these inquiries obscure the most profound metaphysical development of deep learning: the emergence of "latent space." We argue that latent space, the high-dimensional mathematical territory where neural networks process information, functions as a new, non-human ontology of meaning. Within this structure, deep learning (...) effects a transition from classical symbolic reasoning to "vectorial semantics," where concepts are defined not by logical definitions but by geometric proximity and statistical density. The article proceeds in five parts. First, it critiques the demand for "explainability," arguing that the opacity of the black box is an essential property of a system that prioritises high-dimensional correlation over human-readable causation. Second, it defines the metaphysics of the latent world, framing it as a "Statistical Platonism" where vectors function as malleable Forms derived from data. Third, it analyses the training dataset as a "geometric unconscious," suggesting that algorithmic bias is a faithful hermeneutic map of the cultural archive rather than a technical error. Fourth, it explores the "inhuman gaze" of the model, an epistemology of pure correlation that challenges the scientific method's reliance on theory and causality. Finally, the article examines the "performative loop" created by generative AI. As synthetic data saturates the internet, the latent map begins to precede and engender the territory of the real, threatening an "ontological collapse" of human culture into a statistical average. We conclude that the philosophical task of the twenty-first century is not merely the alignment of AI with human values, but the cartography of this new, transformative latent reality. (shrink)

Deep learning has emerged as a transformative technology, enabling advancements in fields such as computer vision, natural language processing, and autonomous systems. Python, with its comprehensive libraries and frameworks, has become the primary language for developing deep learning models. This paper explores the latest advancements in Python-based deep learning, focusing on key frameworks, algorithms, and innovations. It also discusses the challenges associated with implementing deep learning solutions, such as computational cost, data quality, (...) and model interpretability. Furthermore, it addresses sustainable approaches to deep learning, emphasizing energy-efficient techniques, model optimization, and the adoption of green computing practices. By understanding these advancements and challenges, we can push forward towards more efficient and sustainable deep learning solutions using Python. (shrink)

Endowing artificial systems with explanatory capacities about the reasons guiding their decisions, represents a crucial challenge and research objective in the current fields of Artificial Intelligence (AI) and Computational Cognitive Science [Langley et al., 2017]. Current mainstream AI systems, in fact, despite the enormous progresses reached in specific tasks, mostly fail to provide a transparent account of the reasons determining their behavior (both in cases of a successful or unsuccessful output). This is due to the fact that the classical problem (...) of opacity in artificial neural networks (ANNs) explodes with the adoption of current Deep Learning techniques [LeCun, Bengio, Hinton, 2015]. In this paper we argue that the explanatory deficit of such techniques represents an important problem, that limits their adoption in the cognitive modelling and computational cognitive science arena. In particular we will show how the current attempts of providing explanations of the deep nets behaviour (see e.g. [Ritter et al. 2017] are not satisfactory. As a possibile way out to this problem, we present two different research strategies. The first strategy aims at dealing with the opacity problem by providing a more abstract interpretation of neural mechanisms and representations. This approach is adopted, for example, by the biologically inspired SPAUN architecture [Eliasmith et al., 2012] and by other proposals suggesting, for example, the interpretation of neural networks in terms of the Conceptual Spaces framework [Gärdenfors 2000, Lieto, Chella and Frixione, 2017]. All such proposals presuppose that the neural level of representation can be considered somehow irrelevant for attacking the problem of explanation [Lieto, Lebiere and Oltramari, 2017]. In our opinion, pursuing this research direction can still preserve the use of deep learning techniques in artificial cognitive models provided that novel and additional results in terms of “transparency” are obtained. The second strategy is somehow at odds with respect to the previous one and tries to address the explanatory issue by avoiding to directly solve the “opacity” problem. In this case, the idea is that one of resorting to pre-compiled plausible explanatory models of the word used in combination with deep-nets (see e.g. [Augello et al. 2017]). We argue that this research agenda, even if does not directly fits the explanatory needs of Computational Cognitive Science, can still be useful to provide results in the area of applied AI aiming at shedding light on the models of interaction between low level and high level tasks (e.g. between perceptual categorization and explanantion) in artificial systems. (shrink)

Coronavirus disease has a crisis with high spread throughout the world during the COVID19 pandemic period. This disease can be easily spread to a group of people and increase the spread. Since it is a worldly disease and not plenty of vaccines available, social distancing is the only best approach to defend against the pandemic situation. All the affected countries' governments declared locked-down to implement social distancing. This social separation and persons not being in a mass group can slow down (...) the spread of COVID19. It reduces the physical contact between infected persons and normal healthy persons. Almost every health organization tells that to follow social distancing people should maintain at least 6 feet of distance from each other. This research proposes a deep learning approach for social distancing which is developed for tracking and detecting people who are in indoor as well as outdoor scenarios using YOLO V3 video analytic technique. This approach focuses to inspect whether the people are maintaining social distancing in many areas, using surveillance video with measuring the distance in real-time performance. Most of the early studies of detecting social distance monitoring were based on GPS for tracking the movements of people where the signals could be lost. On the other hand, some countries use drones to detect large gatherings of people who cannot have a clear view at night times [10]. In the future, the proposed system can be used fully for detecting threats in the public crowded or it can detect any person affected by critical situations (ie fainting, Cordia arrest) or planting the crops in the forms evenly with a uniform measurement. This proposal could be used in many fields like crowd analysis, autonomous vehicles, and human action recognition and could help the government authorities to redesign the public place layout and take precautionary action in the risk zones. This system analyses the social distancing of people by calculating the distance between people to slow downing the spread of the COVID 19 virus. (shrink)

Web attacks, such as SQL injection, cross-site scripting (XSS), and distributed denial-ofservice (DDoS), pose significant threats to the security and integrity of online systems. Traditional detection methods, relying on rulebased systems or shallow machine learning, often struggle to keep pace with the evolving sophistication of these attacks. This paper proposes an end-to-end deep learning framework for detecting web attacks, leveraging the power of neural networks to automatically learn complex patterns and features from raw web traffic data. Unlike (...) conventional approaches that require extensive manual feature engineering, our method processes unprocessed HTTP request logs and network traffic directly, enabling a fully automated detection pipeline. The proposed model employs a combination of convolutional neural networks (CNNs) and recurrent neural networks (RNNs) to capture both spatial and temporal dependencies in web traffic data. CNN layers extract local patterns, such as malicious payloads or anomalous headers, while RNN layers, enhanced with attention mechanisms, model sequential behaviors indicative of attacks. The framework is trained endto-end on a diverse dataset comprising benign traffic and real-world attack samples, ensuring robustness across various attack vectors. Preprocessing is minimized to tokenization and normalization, preserving the raw data's integrity and reducing human intervention. Evaluation results demonstrate that the model achieves superior performance compared to traditional methods, with an accuracy of over 95% and a low false-positive rate. It effectively identifies both known and zero-day attacks by generalizing learned representations. Additionally, the system scales efficiently, processing large volumes of traffic in real-time, making it suitable for deployment in production environments. This end-to-end deep learning approach not only enhances detection accuracy but also adapts to emerging threats, offering a proactive solution to the growing challenge of web security. Future work will focus on integrating explainability to improve trust and interpretability in security operations. (shrink)

Hyperparameter optimization is a crucial aspect of deep learning, as the choice of hyperparameters significantly influences model performance. Finding the optimal set of hyperparameters can be a time-consuming and computationally expensive process. Traditional techniques, such as grid search and random search, often fail to efficiently explore the vast hyperparameter space, especially for deep learning models with numerous parameters. In this paper, we propose Bayesian Optimization (BO) as an effective approach for hyperparameter optimization in deep (...) class='Hi'>learning models. Bayesian Optimization is a global optimization technique that is particularly suitable for optimizing complex, expensive to-evaluate functions. Unlike grid search or random search, BO builds a probabilistic model of the objective function and uses this model to make informed decisions about where to search next in the hyperparameter space. This approach reduces the number of evaluations required to find optimal or near-optimal hyperparameters, making it computationally efficient and well-suited for deep learning applications. The paper presents a detailed overview of Bayesian Optimization, its working principles, and how it can be applied to deep learning hyperparameter tuning. We explore the use of Gaussian Processes (GP) as surrogate models for BO and highlight the benefits of using acquisition functions to balance exploration and exploitation. Additionally, we compare BO with traditional methods, evaluating its performance in various deep learning tasks such as image classification, natural language processing, and time-series forecasting. Finally, we discuss the challenges and limitations of using Bayesian Optimization for hyperparameter tuning and offer insights into future directions for improving its efficiency and applicability in large-scale deep learning models. (shrink)

there is a communication gap between hearing-impaired people and those with normal hearing, sign language is the main means of communication in the hearing-impaired population. Continuous sign language recognition, which can close the communication gap, is a difficult task since the ordered annotations are weakly supervised and there is no frame-level label. To solve this issue, we compare the accuracy of each model using two deep learning models, Inception and Xception. To that end, the purpose of this paper (...) is By attempting to translate sign language using deep learning models, we focused in this paper a comparing between Inception and Xception ¬¬¬¬¬ algorithm and compared it to previous papers in which we used VGG19[1], mobileNet [2], where we get in the Inception and Xception algorithm on the degree of Accuracy 99.32%, 99,43% respectively. number of images (29000 in the dataset in size 75*75 pixel ) and the data split training data into training dataset (70%) and validation dataset(15%) and testing dataset(15%) and 20 epoch. (shrink)

This work demonstrates the implementation and use of an encoder-decoder model to perform a many-to-many mapping of video data to text captions. The many-to-many mapping occurs via an input temporal sequence of video frames to an output sequence of words to form a caption sentence. Data preprocessing, model construction, and model training are discussed. Caption correctness is evaluated using 2-gram BLEU scores across the different splits of the dataset. Specific examples of output captions were shown to demonstrate model generality over (...) the video temporal dimension. Predicted captions were shown to generalize over video action, even in instances where the video scene changed dramatically. Model architecture changes are discussed to improve sentence grammar and correctness. (shrink)

This work demonstrates the implementation and use of an encoder-decoder model to perform a many-to-many mapping of video data to text captions. The many-to-many mapping occurs via an input temporal sequence of video frames to an output sequence of words to form a caption sentence. Data preprocessing, model construction, and model training are discussed. Caption correctness is evaluated using 2-gram BLEU scores across the different splits of the dataset. Specific examples of output captions were shown to demonstrate model generality over (...) the video temporal dimension. Predicted captions were shown to generalize over video action, even in instances where the video scene changed dramatically. Model architecture changes are discussed to improve sentence grammar and correctness. (shrink)

Die aktuellen Erfolge von Künstlicher Intelligenz beruhen nicht nur auf technologischen Fortschritten, sondern auch auf einem grundlegenden soziotechnischen Strukturwandel. Denn maschinelle Lernverfahren wie Deep Learning benötigen eine große Menge Trainingsdaten, die nur über menschliche Mitarbeit gewonnen werden können. In einer Konvergenz von Methoden der Human-Computer-Interaction und der KI ist in den letzten zehn Jahren eine Fülle von Mensch-Maschine-Interfaces und medialen Infrastrukturen entstanden, durch die menschliche kognitive Ressourcen in hybride Mensch-Maschine-Apparate eingespannt werden. Diese Apparate vollbringen im Ganzen jene Leistung, (...) die als künstliche Intelligenz bezeichnet wird. KI in der damit vorliegenden Form ist somit ein distribuiertes, netzwerkförmiges Phänomen. (shrink)

This study investigated the comparative impact of chat-bots and deep learning of accounting officers' work effectiveness in Anambra State. A correlation research design was adopted, and a structured questionnaire was administered using Artificial Intelligence Questionnaire (AIQ) and Work Effectiveness Questionnaire (WEQ) to 221 accountants in business organizations. The study found a moderate positive relationship between automated chat-bots and work effectiveness (r = .465, N = 221) and a low positive relationship between deep learning machines and work (...) effectiveness (r = .305, N = 221). The study also revealed significant relationships between both automated chat-bots and deep learning machines with work effectiveness. The findings have implications for business organizations, highlighting the need to adopt AI technologies to enhance financial management capabilities, productivity, and efficiency. The study recommends that business organizations adopt automated chat-bots and deep learning machines to improve accounting practices and provide training and development opportunities for accounting officers. This study contributes to knowledge theoretically, practically, and methodologically, providing insights into the relationship between AI technologies and work effectiveness in accounting. (shrink)