Researchers from BRAC University and SINTEF Digital have introduced a groundbreaking hybrid AI model to detect stalking in real time using video footage. This advanced system combines CNN, LSTM, and MLP technologies to analyze facial landmarks, head movements, and spatial distances, achieving an impressive 89.58% accuracy. The approach addresses the alarming issue of physical stalking, often a precursor to serious crimes like assault or kidnapping. The model classifies stalking vs. non-stalking scenarios using minimal video frames. This innovation also introduces the first publicly available dataset featuring stalking and non-stalking incidents, paving the way for further research.

Read the full paper here: https://arxiv.org/html/2402.03417v1#S5

#AI #StalkingDetection #DeepLearning #ComputerVision #PublicSafety #SurveillanceTech #Innovation #WomenSafety #MLP #CNN #LSTM