The Impact of Image Resolution in the training of Generative Adversarial Networks for Violence Detection

Abstract

Since time immemorial, violence has been a problem that the world has been facing. The rise of technology has presented an opportunity to help in this matter. Violence detection solutions have been created for this cause. The problem with existing solutions is that they are not appropriate for settings in a developing country. Factors such as the place, objects seen, people involved, among others, are different from those models who are trained with datasets from developed countries, which might prove ineffective for developing countries. That is why the researchers aim to create a Generative Adversarial Networks Model trained with data that are location-specific to the country of Philippines. In this study, the researchers will gauge the effects and impact that resolution brings in the training of the GAN Model, named V.GAN, to help with improving its performance and implementation.