Hybrid model combining CNN and LSTM achieves 93.90% accuracy in detecting interpersonal violence in videos, suggesting improved performance over existing methods.
It is common that interpersonal violence is recurrent in public spaces, these are manifested in different ways such as punching, slapping, kicking and pushing, being recorded by video surveillance cameras, these records of images are currently processed by algorithms that are able to detect interpersonal violence, but it is necessary to further improve performance. This paper proposes a hybrid model combining Xception Convolutional Neural Networks (CNNs) and Long Short-Term Memory (LSTM) for the detection of violence in videos. We evaluate the effectiveness of our proposal using two datasets: the Hockey fight dataset and real life violence situations dataset. The results showed an accuracy of 93.90% and 98.50% respectively, highlighting that the best performance was achieved with the real life violence situations dataset, comparing the proposed hybrid model with other models proposed in related work, the one we propose shows better performance
No takes yet. Share an insight, caveat, or question.
Calderon-Vilca et al. (2025) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: