Automated Image Forgery Detection through Classification of JPEG Ghosts
Fast start.
Try code below and you will see the magic:
from manage import *
path = 'data/trump-queen-fake.jpg'
predict_img(path, w=16)
- In console
git clone https://github.com/kalinkinisaac/auto-forgery-detection.git
- Set up your virtual enviroment (python3)
- Install packages
pip install -r requirements.txt
- Go to the folder
cd auto-forgery-detection/
- You are ready to go! You are perfect!
- Open python console, when you are in project folder
python
- Import all from managy.py
from manage import *
- Let be you photo you want to analise be
path = /Users/user/Downloads/test.jpg
- Determine path to photo in console
path = 'path'
- Run prediction script
predict_img(path, w=64)
- You will see your image, and the prediction process will start.
- If your image analisys is not accurate, than you should decrease value of
w
. For example here you can see differencew=64
to the right, andw=16
to the left: - After you guessed the right
w=w0
, than you are able to predict finally.
Green and Red squares (named blocks) represents difference between to images.
- Article: 'Automated Image Forgery Detection through Classification'
- Documentation of scikit-learn 0.21.2
- NumPy User Guide
- Kalinkin Isaac (@kalinkinisaac)
- Nikolenko Alex (@GorillazZz1)
- Novik Ivaan