Recognition of Potentially Dangerous Selfies in Real Life

Capturing and posting the images of a selfie has become an ongoing trend in recent years. Since March 2016, there has been the advent of vulnerable events such as the death of people in large numbers and huge groups were injured while clicking selfies. Researchers have studied selfies for understanding the psychology of the authors and understanding their effect on social media platforms. This research work includes a detailed analysis of the casualties related to selfies, and the reason behind the dreadful incidents is considered. Image type is used to classify a particular selfie as a dangerous of which type. The image types were chosen because they contained detailed information that enabled the prediction. Various methods were implemented on 130 selfies that were annotated were collected from Twitter, giving a 58.65% accuracy overall. The collected images have their background with Water, Animal, and Gun. These have been the key factors by which images were classified using them as dangerous. Four types of Classifiers were used so that the accuracy could be estimated in an optimal manner. The classifier algorithms that have been implemented in this research work are Simple Virtual Machine (SVM), Decision Tree, NaiveBayes, and K Nearest Neighbor (KNN).