Detecting and Locating Image Forgeries with Deep Learning
Kamred Udham Singh, Ankur Rao, Ankit Kumar, Neeraj Varshney, Pushpendra Singh Chundawat, Teekam Singh · 2024
It is becoming more important to have reliable methods for identifying picture forgeries, particularly passive forging techniques like copy-move and splicing, since the proliferation of digital images across a variety of platforms has raised the urgency of the situation. The purpose of this research is to explicitly target passive forgery tactics by presenting a unique strategy that makes use of deep learning approaches for the purpose of accurately identifying modified areas inside photographs. The methodology that we have suggested makes use of Mobile Net and ResNet as feature extraction networks. It does this by extracting rich representations from picture data in order to identify complicated patterns that are suggestive of possible manipulations, such as copy-move and splicing manipulations. We make use of Mask R-CNN, which is a cutting-edge instance segmentation model, in order to precisely localize and quantify forged areas. This allows us to calculate the proportion of changed material that is included inside photos.