UCI-SAP-Capstone-2024 / Proof-of-Performance-Fraud-Detection

0 stars 1 forks source link

Image Manipulation model #22

Open NeilNagaraj opened 7 months ago

NeilNagaraj commented 7 months ago

Worked on creating a model to detect image manipulation. Started off using the PS-Battles dataset along with the CNN model in this article (https://www.javatpoint.com/image-forgery-detection-using-machine-learning). The accuracy here was pretty low and ranged around 60-70%. Then I switched to the CASIA-2 dataset (https://www.kaggle.com/datasets/jayaprakashpondy/casia2-dataset) that contained training as well as test images. Ran a few models like https://github.com/agusgun/FakeImageDetector, until I was finally able to detect manipulated images with a CNN model that gave an accuracy of 80-90% atleast.