The increasing sophistication of deepfake technology poses a substantial risk to biometric security, particularly in palmprint recognition systems. As palmprint authentication gains widespread adoption in financial transactions and identity verification, the potential for deepfake-based spoofing attacks becomes a major concern. Fraudsters can manipulate or synthetically generate palmprint images to bypass authentication systems, leading to financial fraud and compromised security. This research focuses on the development of an effective deepfake palmprint detection framework leveraging deep learning techniques. A combination of Convolutional Neural Networks (CNNs) and advanced architectures such as ResNet and custom-designed models is employed to extract distinguishing features and enhance classification accuracy. To improve model robustness, a unique dataset is created by generating deepfake palmprints using ridge extraction and blending techniques, followed by augmentation with transformations like contrast variation, random rotation, perspective distortion, and noise addition. The dataset is systematically split for training, validation, and testing to ensure comprehensive evaluation. Performance metrics including accuracy, precision, recall, and F1-score are used to assess model effectiveness, with a focus on achieving high detection rates. This research contributes to strengthening biometric security by providing a reliable solution to detect deepfake palmprints, particularly in high-risk applications such as banking and secure authentication systems.
Vision impairment is one of the top ten disabilities that can be affected by humans, with India having the highest visually impaired population. This study introduces a novel framework utilizing YOLOv11 for real-time object detection and recognition for the objects and to assist the visually impaired individuals in navigating their surroundings independently. The YOLOv11 detector, trained on the COCO dataset with an additional class for enhanced detection, enables precise classification of various objects. The system integrates Python and OpenCV to detect objects from videos and webcams, ultimately providing audio feedback to the user. The proposed system ensures fast and accurate detection, improving accessibility and reducing dependency on external assistance.
Triple riding with non-helmet usage detection stands as a vital practice for road security enforcement. Through deep learning and computer vision technology automated surveillance systems now perform effective tracking of road safety violations. The presented system incorporates YOLO-based object detection with PaddleOCR license plate recognition to perform its AI- driven operations. The system runs real-time tracking of helmets along with triple riding behaviours before storing violation data in an SQLite database for examination purposes. Experimentation revealed that the object detection capability reached 92% accuracy while OCR- based license plate recognition accuracy amounted to 88%. The system contains an automatic alert system which supports enforcement agencies. Through its application framework the proposed system improves road safety by enhancing law enforcement operations while cutting down manual work processes.
Digital image forgery has emerged as a critical challenge in the digital era, with deep learning techniques offering promising solutions for detecting manipulated images. This paper presents a comprehensive approach to image forgery detection using convolutional neural networks (CNNs) and deep learning models such as ResNet, EfficientNet, and Vision Transformers (ViTs). The methodology involves dataset preprocessing, feature extraction, and classification of forged versus authentic images. Performance evaluation metrics such as accuracy, precision, recall, and F1-score are used to compare different models. The proposed framework enhances the reliability of digital image authentication, ensuring robust forensic analysis for cybersecurity, media integrity, and legal evidence verification. Experimental results demonstrate the efficacy of the deep learning-based approach in detecting various forgery types, including copy-move, splicing, and GAN-generated fake images.
Flight delays are one of the most significant issues in the aviation sector, resulting in money-related calamities for planes and inconvenience for passengers. with millions of flights operating annually, increasing discuss activity has led to congestion and wasteful operational aspects. delays are influenced by various variables, including climate conditions, discuss activity, specialized issues, and airplane terminal operations. in order to solve this problem, we introduce a deep learning-based demonstrate for flight delay expectation based on verifiable and real-time data. our strategy involves examining crucial variables like flight plans, climate designs, and flying machine operations. We will implement deep learning calculations, including LSTM and RNN, to separate designs and predict delays accurately. The think about refers to the comparison of the performance of different models to determine the most effective approach. the results of this research can assist carriers in making informed decisions, reducing delays, lowering costs, and enhancing traveler satisfaction through much improved operation management and planning.
