Published Paper Details:

The rising frequency of violence against women in public spaces demands the design of sophisticated early intervention systems capable of detecting and mitigating threats in real time. This study introduces WATCHGUARD, a novel surveillance platform powered by artificial intelligence, which merges cutting-edge computer vision techniques and deep learning to enhance the safety of women in urban surroundings. The architecture adopts a multi-modal design: YOLOv8 performs object detection, the OpenAI CLIP model conducts semantic validation of actions, and a suite of custom neural networks executes gesture recognition and gender classification. Our experimental framework analyzes continuous video streams to detect dangerous objects, recognize distress signals, and automate alerts to law enforcement. The prototype achieves 94.2% accuracy in overall threat identification, 91.8% precision in gender classification, and 89.5% performance in gesture recognition, while sustaining a low false-positive rate of 3.2%. Validation across varied urban contexts demonstrates the ability to recognize threat manifestations, including physical assaults, weapon-mediated attacks, and standardized emergency gestures. Privacy is safeguarded through robust data anonymization protocols, and the architecture is designed for seamless integration within existing smart city infrastructures. The findings demonstrate substantial decreases in response times accompanied by heightened situational awareness among security staff, establishing this system as a feasible and proactive measure for advancing women's safety in urban environments.