A hybrid optimization-tuned deep convolutional neural network for bare skinned image classification in websites

With the advent of the Internet, social media, and mobile technologies, pornographic images have been broadly disseminated and caused great destruction to the social stability and the psychology of adolescents. Furthermore, pornographic content acts as one of the major causes of crimes and abuses, and hence it is crucial to identify such images in the websites. This paper proposes an optimization tuned Deep Convolutional neural network (Deep-CNN) model for classifying pornographic images in websites. The significance in the classification of such images relies on the utilization of the proposed spotted hyena Aquila (SHyAq) optimization algorithm that inherits the characteristics of the hyena hunters and the Aquila hunters in tuning the tunable weights of the Deep-CNN model optimally. In addition, the performance of the proposed SHyAq-based Deep-CNN model is enhanced using the significant features of the image extracted using the feature extraction strategy. Finally, the proposed porn image classification model analysis is carried out based on the performance metrics, such as accuracy, sensitivity, and specificity. The results thus obtained are compared with the existing methods to validate the effectiveness of the proposed model in porn image classification. The proposed SHyAq-based Deep-CNN technique outperformed other states of the art techniques like AIRNet, Multiple feature fusion transfer learning, MLP, FSVM, DOCAPorn, CNN, Multi-level CNN, Deep CNN, Aquila-based Deep CNN, Coyote-based Deep CNN, and AqCO-based Deep CNN in terms of accuracy, sensitivity, and the specificity with the values of 96.46% each, respectively.