An Enhanced Multimodal Stacking Scheme for Online Pornographic Content Detection

Abstract

An enhanced multimodal stacking scheme is proposed for quick and accurate online detection of harmful pornographic contents on the Internet. To accurately detect harmful contents, the implicative visual features (auditory features) are extracted using a bi-directional RNN (recurrent neural network) with VGG-16 (a multilayered dilated convolutional network) to implicitly express the signal change patterns over time within each input. Using only the implicative visual and auditory features, a video classifier and an audio classifier are trained, respectively. By using both features together, one fusion classifier is also trained. Then, these three component classifiers are stacked in the enhanced ensemble scheme to reduce the false negative errors in a serial order of the fusion classifier, video classifier, and audio classifier for a quick online detection. The proposed multimodal stacking scheme yields an improved true positive rate of 95.40% and a false negative rate of 4.60%, which are superior values to previous studies. In addition, the proposed stacking scheme can accurately detect harmful contents up to 74.58% and an average rate of 62.16% faster than the previous stacking scheme. Therefore, the proposed enhanced multimodal stacking scheme can be used to quickly and accurately filter out harmful contents in the online environments.

Multimodal music emotion recognition in Indonesian songs based on CNN-LSTM, XLNet transformers

“…These techniques aim to apply the output of a multiclass model to generate a new model, this method is seen in these studies [27]- [32]. The ensemble method acts as a meta or sub-classifier for the basic classifier's output, where the basic classifier's output has been trained for different tasks or features [33], [34]. The stacking method has achieved excellent performance for image and text classification tasks where these statements also fulfill their task [9].…”

Obscene image detection using transfer learning and feature fusion

“…In this subsection we will compare performance behavior of our proposed TLMoBDense model with some of the state-of-the-art methods [ 4 , 5 , 7 , 9 , 16 , 19 , 25 , 26 , 29 , 32 ] in literature. Table 9 provides comparison of classification quantifying parameters of these methods for different dataset images.…”

“…Authors in [ 25 ] used Inception V3 model to address the sexually explicit videos detection problem. In [ 29 ], authors integrated multi-modal stacking strategy for the detection of porn contents in the social media via online mode. Authors have fused bidirectional recurrent neural network and VGG 16 model to achieve an accuracy of 95.60 % .…”

“…In literature, there exists varieties of obscene image detection methodologies (OIDMs) for identification and detection of obscene contents and consequently enable to prohibit/block their further use at the user end. These methodologies are broadly categorized as i) skin-color segmentation based methods [ 5 , 8 , 16 ] ii) Bag of words (BoWs) based models [ 6 , 14 , 18 ] iii) visual and motion feature analysis based [ 11 , 19 , 21 , 30 , 31 ] and iv) deep learning-based classification strategies [ 1 , 4 , 7 , 9 , 22 , 24 – 27 , 29 , 32 , 33 ]. Usually, skin-color segmentation based methods are highly dependent on human skin pixels and hence, unable to generalize the findings.…”

A Deep Learning Approach to Detect Pornography Videos in Educational Repositories

“…al. [Song and Kim 2020] proposed a multimodal stacking scheme for quick and accurate online detection of pornographic content. Their work uses both visual and auditory features as input for their detection method.…”