Image Forgery Detection with ELA and Deep Learning

Project objective

Combine the implementation of error-level analysis (ELA) and deep learning to detect whether an image has undergone fabrication or/and editing process or not, e.g. splicing.

Methods

1. Error-level analysis
2. Convolutional neural networks (CNN)

---

Activity Diagram

Architecture

Result

• Convergence: Epoch 30
• Best accuracy: 98.13% (epoch 30)