Final Project for Deep Learning (Fall 2019)@ GWU
Oct. 2019

Authors: Chirag Jhamb, Sean Pili and Poornima Joshi

Most malicious photo manipulations are created using standard image editing tools, such as Adobe Photoshop. We present a method for detecting such manipulations using several pretrained neural networks.The aim of this project is to use Neural Networks to determine whether a photo has been altered by photoshop. During the course of this project, we have attempted to explore several different types of neural networks (like Dilated Residual Networks, ResNets, VGG’s etc,) and frameworks (pytorch, keras, tensorflow). In addition we also experimented with fastai library is order to speed up the training. The main focus throughout has been to achieve better accuracy in identifying photoshopped images. We attempted to isolate the faces in the images in order to reduce the noise in the background. This research is part of a broader effort to improve digital image forensics; a field dedicated to detecting the authenticity of images.

Data was obtained from Kaggle Navigate into the Code folder to know more on running the models.

Some Sample images are seen here:

Levels in the images