Volume 12 | Issue 5 |

 Abstract

Stocksage is a transformative project aimed at revolutionizing stock market participation. By tackling barriers like financial literacy and market complexities, it empowers users with cutting-edge tools. The platform integrates Web Development upon machine learning and regression algorithms for predictive insights, fosters financial literacy, and promotes inclusive and informed investment decisions, both nationally and globally. Beyond its technical sophistication, Stocksage aspires to be an educational resource, promoting financial literacy and understanding of stock market dynamics. By addressing the multifaceted challenges faced by investors, Stocksage endeavors to unlock the latent potential of the stock market, empowering individuals to make informed decisions and participate actively in wealth creation.

Licence: creative commons attribution 4.0

  License

Creative Commons Attribution 4.0 and The Open Definition

 Keywords

Stock Market, Financial Literacy, Web Development, Machine Learning, Deep Learning

  License

Creative Commons Attribution 4.0 and The Open Definition

 Abstract

Distinguishing between authentic and fraudulent information has grown more challenging in the modern information landscape owing to the quick spread of news via digital channels. This project creates and implements a system for detecting fake news that uses state-of-the-art deep learning techniques, specifically Long Short Term Memory (LSTM) neural networks, to address this issue

Licence: creative commons attribution 4.0

  License

Creative Commons Attribution 4.0 and The Open Definition

 Keywords

Fake News, LSTM, Deep Learning, Neural Network

  License

Creative Commons Attribution 4.0 and The Open Definition

 Abstract

Citechgram is a social media platform that draws its inspiration from Twitter and it serves Cambridge University students alone. In this research, we will discuss the possibilities of using cloud computing and web development technologies in building Citechgram. Citechgram can provide an adaptable, safe, and dynamic platform for Cambridge students to connect with each other, share ideas and create a vibrant online community by exploiting cloud-based infrastructure and strong web development frameworks.

Licence: creative commons attribution 4.0

  License

Creative Commons Attribution 4.0 and The Open Definition

 Keywords

cloud services, Web Application, CITECHGRAM.

  License

Creative Commons Attribution 4.0 and The Open Definition

 Abstract

This project focuses on developing a dog identification app using deep learning principles, particularly convolutional neural networks (CNN) and the Inception model. It begins with collecting and preprocessing datasets to optimize images for model training. Emphasis is placed on selecting the appropriate model architecture for precise classification. Mechanisms for detecting dogs in user-supplied images are implemented, and procedures for saving and deploying trained models through an API are established. The app aims to provide accurate identification and classification of dog breeds, catering to researchers studying both physical and behavioral traits. This deep learning-powered solution offers a practical means for canine breed classification, leveraging state-of-the-art image processing techniques.

Licence: creative commons attribution 4.0

  License

Creative Commons Attribution 4.0 and The Open Definition

 Keywords

Dog Identification, Deep Learning, Image Processing, CNN.

  License

Creative Commons Attribution 4.0 and The Open Definition

 Abstract

In computer vision, the field of face emotion for circumstances that comprises a pre-processing detection is expanding with the goal of identifying and comprehending human emotions from facial expressions. We provide a system in this project that combines the CNN (Convolutional Neural Network) and Haar Cascade classification techniques. A sizable dataset of labeled facial expressions is used to train the CNN model, which 21 then uses these characteristics and patterns to identify various emotions. It allows for accurate emotion categorization by extracting high-level abstract features from the input photos. However, the Haar Cascade classifier adds more data for emotional analysis by identifying face landmarks like the lips, nose, and eyes. This cyclical method makes it easier to analyze the emotional states.

Licence: creative commons attribution 4.0

  License

Creative Commons Attribution 4.0 and The Open Definition

 Keywords

Convolutional Neural Network, Haar Cascade Classifier, Facial Emotion.

  License

Creative Commons Attribution 4.0 and The Open Definition

 Abstract

Convolutional Neural Networks (CNNs) play major role in accurately classifying brain tumors identified in medical scans such as MRI. This study presents a CNN architecture tailored specific task particular objective, contains convolutional layers to extract features followed by maximum pooling layers designed for dimensionality reduction. To prevent excessive fitting, dropout layers are employed strategically integrated, ensuring the generalizability of the model fitting. The task of classification involve using fully connected layers with the Softmax The activation function utilized in the suggested CNN architecture demonstrates effectiveness in Classifying brain tumors into three categories distinct types: meningioma, glioma, and pituitary tumors. Experimental evaluation reveals promising results, with the model achieving an overall classification accuracy of 98%. Specifically, it detects glioma with 96% accuracy, identifies no tumor with 99% accuracy, differentiates meningioma with 97% accuracy, and identifies pituitary tumors with 99% accuracy. The dataset comprises 3264 images, 90% of which are for training and 10% for testing. This method holds considerable potential to assist clinicians in accurate and timely diagnosis, thereby facilitating suitable treatment planning for patients with brain tumors. Further research can explore improvements to the network architecture and explore its applicability in different medical imaging datasets.

Licence: creative commons attribution 4.0

  License

Creative Commons Attribution 4.0 and The Open Definition

 Keywords

CNN, maximum pooling layers, dropout layers, softmax activation

  License

Creative Commons Attribution 4.0 and The Open Definition

 Abstract

This project presents an innovative application designed for both standalone and interconnected frameworks, aimed at the real-time automatic detection and localization of food items within dynamic scenes. By harnessing diverse configurations such as Single Shot Detection, Faster R-CNN, YOLO, EfficientDet, RetinaNet, and Mask R-CNN, this study employed A thoroughly managed dataset derived from various online repositories. These configurations were seamlessly integrated with food detection models and multiple convolutional network architectures, harnessing the power of multiple neural networks to enhance performance. Computer vision, an integral part of artificial intelligence, is employed to replicate human perception of three- dimensional structures in visual environments. Through the amalgamation of digital images and advanced deep learning models, this research aims to enable computers to interpret and comprehend the visual world, specifically focusing on the identification and classification of food items. Within the framework of the food industry, this paper highlights the significance of precise object recognition and classification, crucial for ensuring a nutritious diet and overall well-being. With the burgeoning advancements in nutritional science and the availability of diverse smartphone applications, the research aims to present a comprehensive framework capable of autonomously identifying, categorizing, and localizing food elements in various scenes and settings.

Licence: creative commons attribution 4.0

  License

Creative Commons Attribution 4.0 and The Open Definition

 Keywords

Deep Learning, SSD, EfficientDet, YOLO, Faster R- CNN, RetinaNet, Mask R-CNN

  License

Creative Commons Attribution 4.0 and The Open Definition

 Abstract

Sign language is an essential communication language that helps individuals with hearing loss interact with others. Convolutional Neural Network is a useful tool for image processing tasks, including sign language recognition. This paper proposes a novel CNN-based approach to sign language to text translation. The CNN model is intended to effectively extract temporal and spatial properties from sign language containing video sequences. Convolutional layers are utilized to extract hierarchical features, while pooling layers are employed to decrease spatial dimensions without sacrificing important information. The model in this work is trained on a large dataset of sign language images, allowing strong representation learning for accurate translation. The CNN model has performed well in translating American Sign Language into text, according to test results. Using datasets of American Sign Language, the model surpasses previous methods and reaches high accuracy. In general, the suggested CNN-based approach for translating sign language to text provides a pathway between people who use sign language and others who are not familiar with it. By providing real-time translation capabilities, this tool can improve accessibility and inclusivity for the community in a range of situations, such as everyday communication, healthcare, and education. This research promotes more equality and integration for those with hearing loss and enhances assistive technologies.

Licence: creative commons attribution 4.0

  License

Creative Commons Attribution 4.0 and The Open Definition

 Keywords

Sign language recognition, Image processing, Deaf communication, Gesture-to-text conversion

  License

Creative Commons Attribution 4.0 and The Open Definition

 Abstract

This paper presents a comprehensive Quality Assurance Framework designed specifically for text generation models. Our approach combines automated metrics such as BLEU, ROUGE, and perplexity scores with novel techniques for coherence and factuality assessment. We integrate human evaluation methodologies to ensure a balanced assessment of linguistic quality, coherence, factual accuracy, and diversity in generated texts. Through extensive experimentation across different text generation tasks, our framework demonstrates improved evaluation accuracy and provides valuable insights for model refinement and optimization, contributing to the advancement of trustworthy text generation models.

Licence: creative commons attribution 4.0

  License

Creative Commons Attribution 4.0 and The Open Definition

 Keywords

A Quality Assurance Framework for Evaluation of Text Generation Models

  License

Creative Commons Attribution 4.0 and The Open Definition

 Abstract

Distributed denial of service (DDoS) threats represents a significant cybersecurity challenge, constituting a variant of denial of service (DoS) in which IP addresses are exploited to launch attacks to a specific host or victim. DDoS attacks, characterized by meticulous coordination, exploit compromised secondary victims to target one or more victim systems, ranging from large-scale enterprise servers to less. These threats incur significant bandwidth and power costs, leading to the compromise of confidential data. Therefore, developing advanced algorithms to accurately detect various DDoS cyber threats, while considering computational load, has become urgent. The majority of the research currently in publication approaches DDoS threat detection as a binary classification problem, that is ascertaining whether or not an attack has started. However, to effectively protect the network and minimize significant damage, it's critical to distinguish the specific type of DDoS attack targeting the network or system. This study presents a comprehensive classifier that combines the strengths of the four best-performing algorithms. A comparative analysis is performed, comparing the Classifier with different artificial intelligence and machine learning (AI and ML) algorithms. Its goal is to improve the identification of various kinds of DDoS threats by transforming the problem into a multi-label classification scenario. Through this approach, the research aims to contribute to the refinement of cybersecurity strategies, ensuring a deeper understanding and proactive defence against various DDoS cyber threats.

Licence: creative commons attribution 4.0

  License

Creative Commons Attribution 4.0 and The Open Definition

 Keywords

DDoS, cybersecurity, classification, multi-label, detection, attacks, algorithms, proactive defence, threat identification, network security.

  License

Creative Commons Attribution 4.0 and The Open Definition