Engineering Genesis

Sentinel Pay: A Hybrid AI-Driven System for Real-Time Fraud Detection in Digital Payment Systems

2026-05-05T12:52:04+00:00

Financial system fraud goes undiscovered, even when it would be most beneficial for them to do so. UPI transactions typically complete in less than two seconds, making it nearly impossible to
detect any mistakes after the fact. SentinelPay circumvents this issue by classifying all transactions prior to the payment gateway approval at the authorization step. The architecture was developed using an event-driven microservices approach, and each part was designed to maintain the latency below acceptable boundaries. A few design choices shape the system. Apache Kafka takes care of transaction events to keep submission and scoring from being mixed up. There are various levels in the detection pipeline. After discovering statistical outliers and starting with the basic threshold criterion, Spring AI was used to make decisions based on the situation. There was no layer that was missing because they all depend on each other. This makes sure that the detection process works well and finds any mistakes. Redis may maintain user activity profiles, so the inference engine can acquire data in milliseconds. After looking at 50,000 entries in total, SentinelPay found real UPI fraud trends. We looked at the following numbers: recall (94.6%), detection accuracy (95.2%), F1-score (94.2%), and precision (93.8%). The throughput stays the same at 1,200 TPS, even when there is a lot of traffic. The average end-to-end latency is about 50 ms. This method gives results that are 13.2 points more accurate if you start with a rule. There are two other benefits: the frequency of false positives was down by 65%, and the time it took to get findings went down by 75%. You do not have to change your UPI settings to use SentinelPay.

Hand Sign Detection Using Deep Learning

2026-05-06T06:02:01+00:00

The hand sign detection is an artificial intelligence system that detects and interprets sign language. Computer vision-based machine learning-based gestures. The system captures hand language gestures using computer vision and machine learning techniques. The system captures hand processes the visual data with deep learning models like and moves through cameras. Models such as Convolutional Neural Networks (CNN) and other AI algorithms. The core idea of this project is to close the gap in communication between deaf or mute people, and non-signers, by turning their hand gestures into text or speech in real-time. the system enhances access and inclusion in education, health care, and customer services. The proposed system incorporates gesture recognition, computer vision, and machine learning algorithms to make sure that the sign language is accurately detected and efficiently. This technology supports real -time communication and promotes equal opportunities for hearing -impaired individuals. hms. The main objective of this project is to bridge the communication gap between deaf or mute individuals and non-signers by converting hand gestures into text or speech in real time. The system improves accessibility and inclusivity in areas such as education, healthcare, and customer service. By integrating gesture recognition, computer vision, and machine learning algorithms, the proposed system ensures accurate and efficient sign language detection. This technology supports real -time communication and promotes equal opportunities for hearing -impaired individuals.

Vehicle Counting and Detective System

2026-05-05T12:52:04+00:00

One of the most urgent issues of modern urban settings is traffic congestion, road safety and ineffective traffic control. Real-time and accurate information concerning the flow of the vehicles is important in the planning, monitoring and control of transportation systems.
Conventional vehicle counting techniques like manual inspection and sensor-based systems have drawbacks such as expensive installation, not scalable, environmental sensitivity and not able to provide detailed classification of vehicles.
This project is a Vehicle Count and Detection System with Computer Vision and Deep Learning, which has been implemented in main language Python. The system uses video feeds of surveillance cameras and processes them with Open CV and state-of-the-art object detection models like YOLO (You Only Look Once).
The proposed solution has the potential of identifying, monitoring, classifying, and enumerating various categories of vehicles in real time, such as cars, buses, trucks, and two-wheelers. The system can eliminate the issue of double-counting and is accurate even in traffic jams by using tracking algorithms and region-of-interest (ROI) based counting logic.
The system is a cost-effective, non-invasive, and scalable system that can be used in smart city applications and intelligent transportation systems. Traffic statistics generated can help authorities in managing congestion, planning infrastructure and formulation of policies.

Digital Forgery Detection for Certificates and Documents Using Image Processing and AI

2026-05-05T12:52:05+00:00

In the digital age certificate forgery and other official documents are a major menace to academic, corporate and legal organizations. Some of the traditional manual forms of verification can be inefficient, time consuming and subject to human error. The paper presents a Digital Forgery Detection System that combines Image Processing and Artificial Intelligence (AI) to classify elements that might be manipulated in digitized documents with high accuracy and is automated.
The multi-staged framework used in the system has Optical Character Recognition (OCR) to analyze text and extract patch-level features as a way of identifying structural inconsistencies. The model can detect tampered signatures, logos, and seals specifically by employing Convolutional Neural Networks (CNNs) and feature-matching algorithms. The result is a forensic report that points to the existence of manipulated areas and a conclusive authenticity rating.
The architecture takes advantage of a hybrid architecture in the detection of minute pixel-level anomaly, including copy-move or splicing attacks, which are not always noticeable by the naked eye. Image inpainting can also be combined to enable the system to rebuild original backgrounds which can serve as a base to detect disturbed text or foreign objects. This two-layered verification provides that even advanced forgeries that imitate original fonts or textures are correctly identified as such.
Comparison of the algorithm with a wide range of real and fake certificates shows that the CNN-based classification is more accurate in comparison with traditional threshold-based algorithms. The system can be scaled and be depended upon by institutions to provide an objective, quantified authenticity score which can be used in institutional security. This study fills the void between forensic image and real-life document authentication and provides a strong defense against digital fraud.

AI Chatbot for Research Paper Development Using RNN

2026-05-05T12:52:05+00:00

The chatbots based on Artificial Intelligence (AI) are changing the process of developing a research paper by offering automatic services in different phases of the research paper development process, including topic selection, literature review, drafting, editing, and managing references. The suggested system can improve efficiency, quality of writing, and compliance with academic standards, as it uses Natural Language Processing (NLP) and machine learning methods, specifically Recurrent Neural Networks (RNN). The chatbot helps researchers to come up with ideas, structure, polish arguments, and give real-time feedback on grammar, coherence, and the precision of citations. Nevertheless, issues like data bias, ethical issues, and critical thinking restrictions are still important. Although the AI-based systems might be effective in summarizing and providing recommendations, they might not be as thorough as a human analysis is necessary in the original research. This paper examines the effects of AI chatbots in scholarly writing, their advantages, and disadvantages, and offers feasible solutions on how to incorporate the AI tools without compromising originality and academic integrity. The direction of future work is to improve the AI approach and create ethical guidelines on how it can be used in research settings.

Automated Helmet Detection and E- Challan Generation using YOLOv8

2026-05-05T12:52:05+00:00

One of the primary issues in the contemporary transport system is road safety, especially in those nations where two-wheelers are prevalent. Many accidents are caused by the riders not wearing helmets. The proposed research is an automated helmet detection system and e-challan generation system based on the YOLOv8 deep learning model. The system identifies the violation of helmet through traffic surveillance footage, and automatically identifies vehicle number plates by Optical Character Recognition (OCR). The information of the vehicle extracted is compared with a database to get the owner details and produce an electronic challan. The system suggested can minimize the effort of a manual system and enhance efficiency in the enforcement of traffic laws.
The suggested system will involve the use of YOLOv8 (You Only Look Once version 8) which is the latest and innovative object detection algorithm, to locate the motorcyclists and determine whether they wear helmets or not in real time. The model can identify the riders based on the presence of a helmet on them with a high degree of accuracy and speed by processing live CCTV footage or recorded video streams. The architectural design of YOLOv8 is lightweight and advanced and can be easily used in real-life scenarios, such as dynamic lighting, weather, and traffic conditions.
On detection of a violation, the vehicle number plate is captured by the system and the registration number is extracted with the help of Optical Character Recognition (OCR). This is then cross- tabulated with a vehicle registration database, where the owner will be automatically retrieved. A fine in the form of an e- challan (electronic fine) is then created and delivered to the vehicle owner in the registry via digital means of communication like SMS or email. This is an automated solution that will reduce the human factor in monitoring by traffic officers, minimize human error and ensure that traffic regulations are uniformly enforced. Furthermore, it offers a scalable and effective solution to smart city infrastructure through incorporation of artificial intelligence in traffic management systems.

Multi-Face Detection and Recognition System

2026-05-06T06:05:13+00:00

The use of face recognition technology in modern intelligent systems has emerged as one of the main features owing to its extensive application across several security, surveillance, access control and automated attendance systems. Although classical face recognition techniques have been found sufficient in controlled settings in the presence of a single target, they fail in real world situations where there are multiple faces present with differences in illumination, pose, expression, occlusion, and complexity of the background. This project tackles these issues by suggesting an effective real-time multi-face detection and recognition system that uses deep learning methods.
The suggested system combines a progressive convolutional neural network (CNN)-based face detection models and a deep embedding-based face recognition method like FaceNet and ArcFace. The system can detect several faces on both fixed-image and live video feeds, extract discriminative facial features and positively identify individuals by comparing feature embeddings with a stored facial database. The structure is a modular one composed of image acquisition, preprocessing, face detection, feature extraction, face recognition, and visualization modules, which allows it to be scaled and operate in real-time.
The system also uses preprocessing methods like face alignment, normalization, and patch extraction to make the system robust to recognize faces in difficult conditions. Experiments on publicly available data including Labeled Faces in the Wild (LFW), and a dataset collected by the authors show high recognition accuracy, resistance to variations, and can be run in real-time. The findings confirm the efficiency of the suggested strategy on managing dense and uncontrolled spaces.
Altogether, this project illustrates that the systems of multi-face detection and recognition based on deep learning will be highly superior to traditional ones and will offer a viable and scalable solution to the real-life presence in surveillance, smart classrooms, and access control systems.