DETECTION OF DEEP FAKE VIDEOS USING LONG DISTANCE ATTENTION

Authors

  • Mr.K.LAKSHMANA REDDY Author
  • PULIDINDI HARITHA Author

Keywords:

Malicious uploaders, young demographic, animated cartoon videos

Abstract

The exponential growth of videos on YouTube has attracted billions of viewers among 
which the majority belongs to a young demographic. Malicious uploaders also find 
this platform as an opportunity to spread upsetting visual content, such as using 
animated cartoon videos to share inappropriate content with children. Therefore, an 
automatic real-time video content filtering mechanism is highly suggested to be
integrated into social media platforms. In this study, a novel deep learning-based 
architecture is proposed for the detection and classification of inappropriate content in 
videos. For this, the proposed framework employs an ImageNet pre-trained 
convolutional neural network (CNN) model known as EfficientNet-B7 to extract video 
descriptors, which are then fed to bidirectional long short-term memory (BiLSTM) 
network to learn effective video representations and perform multiclass video 
classification. An attention mechanism is also integrated after BiLSTM to apply 
attention probability distribution in the network. These models are evaluated on a 
manually annotated dataset of 111,156 cartoon clips collected from YouTube videos. 
Experimental results demonstrated that EfficientNet-BiLSTM (accuracy = 95.66%) 
performs better than attention mechanismbased EfficientNet-BiLSTM (accuracy = 
95.30%) framework. Secondly, the traditional machine learning classifiers perform 
relatively poor than deep learning classifiers. Overall, the architecture of EfficientNet 
and BiLSTM with 128 hidden units yielded state-of-the-art performance (f1 score = 
0.9267). Furthermore, the performance comparison against existing state-of-the-art 
approaches verified that BiLSTM on top of CNN captures better contextual informationof video descriptors in network architecture, and hence achieved better results in child 
inappropriate video content detection and classification.

Downloads

Published

04-06-2024

How to Cite

DETECTION OF DEEP FAKE VIDEOS USING LONG DISTANCE ATTENTION. (2024). International Journal of Engineering Research and Science & Technology, 20(2), 1048-1065. https://ijerst.org/index.php/ijerst/article/view/371