AI Violence Detection

Overview

This project is the first machine learning project that I have worked on in Taylor’s University. The idea came from watching a YouTube video on the worst jobs in the world and one of them being the human filters for social media. The social media you see is the social media that has already been sanitized. Behind them are real human beings who would need to go through millions of disturbing images to prevent them from being posted on the web. This project aims to automatically classify violent images so that human moderators do not need to go through the psychological trauma. Many thanks to Dr Markos Zampoglou & Dr Papadopoulos who generously gave the dataset they used in their research for my own model training.

Key Features

• ▸CNN model training
• ▸Hyperparameter fine tuning
• ▸Model evaluation

Gallery

Previous slideNext slide