Fake Image Detector

Image Tampering Detection using ELA and CNN

Members

1. Agus Gunawan
2. Holy Lovenia
3. Adrian Hartanto Pramudita

Indonesian paper/documentation

Get it here!

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)

Architecture

Result

• Convergence: Epoch 9
• Best accuracy: 91.83% (epoch 9)

Dataset

Please refer to issue #1.