Soccer fans often prefer watching summaries of football games due to the significant time commitment required to view an entire match. Traditional manual methods for analyzing and extracting exciting clips are tedious and time consuming. Therefore, automate the process of video analysis and summarization is crucial. This paper presents a novel approach for automated soccer video summarization by classifying soccer events: card, corner, foul, and freekick. We implemented an empirical analysis of a hybrid Convolutional Neural Network (CNN) and Gated Recurrent Unit (GRU) architecture. The proposed CNN-GRU model achieved an outstanding accuracy of 99.3% and a validation accuracy of 95.18%. These results demonstrate the effectiveness of our approach in automated the extraction of important soccer events, offering significant improvements in efficiency and accuracy over traditional methods. This work has broad applications in sports video analysis and accurate generation of game highlights.

Soccer fans often prefer watching summaries of football games due to the significant time commitment required to view an entire match. Traditional manual methods for analyzing and extracting exciting clips are tedious and time consuming. Therefore, automate the process of video analysis and summarization is crucial. This paper presents a novel approach for automated soccer video summarization by classifying soccer events: card, corner, foul, and freekick. We implemented an empirical analysis of a hybrid Convolutional Neural Network (CNN) and Gated Recurrent Unit (GRU) architecture. The proposed CNN-GRU model achieved an outstanding accuracy of 99.3% and a validation accuracy of 95.18%. These results demonstrate the effectiveness of our approach in automated the extraction of important soccer events, offering significant improvements in efficiency and accuracy over traditional methods. This work has broad applications in sports video analysis and accurate generation of game highlights.

Automatic License Plate Recognition (ALPR) in Bangladesh faces challenges due to the complexity of Bangla script and low-resolution CCTV footage. This research introduces a YOLOv8-based deep learning model tailored for Bangladeshi license plates, enhancing plate localization, character segmentation, and Bangla script recognition using Easy-OCR. The model leverages Roboflow for data collection, annotation, and augmentation, training on a dataset of 2600 images captured under diverse conditions, including low resolution, harsh weather, and partial obstructions. The model distinguishes license plates from other rectangular objects on vehicles, achieving 94.8% detection and recognition accuracy. These results demonstrate the system's robustness in real-world scenarios, contributing to improved road safety, traffic management, and law enforcement in Bangladesh, marking a significant advancement in ALPR technology for the region. This research marks a significant advancement in ALPR technology for Bangladesh, contributing to improved road safety, efficient traffic management, and enhanced law enforcement capabilities.

Abstract: Privacy-preserving human activity anomaly detection has become increasingly important in sensitive applications such as video surveillance, healthcare monitoring, and assisted living systems. While human action detection techniques offer substantial benefits for automated video and sensor-based analysis, they also raise privacy concerns when deployed in environments that require confidentiality. This paper introduces Obfuscated Action Detection, an innovative framework incorporating a temporal obfuscation component based on Generative Adversarial Networks (GANs) to anonymize sensor data. By leveraging Deep Neural Networks, this framework ensures both high accuracy in anomaly detection and feasibility for real-time application. Extensive experiments demonstrate the capability of Obfuscated Action Detection to achieve robust privacy protection without compromising detection precision, making it a viable solution for applications that prioritize both privacy and reliability. Additionally, this paper presents an overview of related works, summarizing recent advancements and methodologies in privacy-preserving anomaly detection.Keywords: HAR; Human action recognition; privacy preserving; GAN; generative adversarial network; image segmentation;

Privacy-preserving human action recognition is a crucial area of research, particularly in the context of video surveillance, assisted living systems, and healthcare applications. While human action recognition techniques offer significant benefits for automated video analysis, they also raise concerns about individual privacy when deployed in sensitive environments. This paper introduces, StealthGuard incorporates a temporal privacy-preserving component based on generative adversarial networks (GANs) to obfuscate sensor data, thereby preventing the identification of individual people or their activities. This approach utilises deep neural network, ensuring both accuracy in action recognition and real-time deployment feasibility. Through extensive experimental results, StealthGuard demonstrates its ability to achieve high levels of privacy protection while maintaining recognition accuracy making it a promising solution for applications where privacy is paramount. This paper also provides a related works in the field, highlighting approaches and techniques for privacy-preserving human action recognition.

Privacy-preserving human activity anomaly detection has become critical in privacy-sensitive fields like video surveillance, healthcare, and assisted living. While human action recognition offers significant advantages in automated analysis, it raises confidentiality concerns. This paper introduces Obfuscated Action Detection, a novel framework that uses Generative Adversarial Networks (GANs) for temporal obfuscation, ensuring privacy while maintaining high accuracy. By integrating Deep Neural Networks, the framework delivers robust anomaly detection with real-time feasibility. Tested on the UCF101 dataset, the model achieves high accuracy (98.59% to 100%) and strong generalization to unseen data (88.44% test accuracy). With impressive precision (98.14%), recall (99.56%), and F1 score (98.84%), Obfuscated Action Detection effectively balances privacy with performance. The framework shows promise for real-world applications in privacy-critical domains, offering robust privacy protection without compromising detection accuracy. Extensive experiments demonstrate the capability of Obfuscated Action Detection to achieve robust privacy protection without compromising detection precision, making it a viable solution for applications that prioritize both privacy and reliability. Additionally, this paper presents an overview of related works, summarizing recent advancements and methodologies in privacy-preserving anomaly detection.