Detection of Violent Content in Videos Using 2D Attention-Augmented Convolutional Networks and Gated Recurrent Unit
Bhavyesh Sajja, Anurag Kumar Singh · 2023
Nowadays, most online videos over the Internet contain violent and explicit content. These videos are available to anyone. Such content must be moderated as it can be disturbing to a majority of viewers and can negatively impact people of younger generations. Currently, researchers use LSTM and Transformer-based architectures to detect violent content, which requires high computing power that edge devices cannot offer. Hence, in this paper, a resource-efficient, attention-based deep learning technique is developed to identify and handle violent video content. The proposed model has been compared against the state-of-the-art models. Our model shows appreciable improvement in results with minimal computing power.