<h1 align="center">
<a href="https://www.linkedin.com/in/hxazem/"><img src="https://user-images.githubusercontent.com/36288517/70856230-91739480-1ee0-11ea-9d85-6acd691642ab.png" alt="IFD"></a>
</h1>
[important Note] this was my graduation project if you want to reuse it use only training py Files for training Models the whole application will not work as this was for local testing only and i didn't upload h5 files which is trained models because it was about ~3GB size which is too much you can use Kaggle API with collab to train models easily, thanks :)
# Image-Forgery-Detection-using-Deep-learning
Image processing with convolutional neural network to detect tampering in image
## Project Description
This project combines different deep learning techniques and image processing techniques to detect image tampering "Copy Move and Splicing" forgery in different image formats (either lossy or lossless formats). We implement two different techniques to detect tampering. I built my own model with ELA preprocessing and used fine tuning with two different pre-trained Models (VGG19 , VGG15) which are trained using [Google Colab](https://colab.research.google.com/notebooks/welcome.ipynb#recent=true),Image Forgery detection application gives user the ability to test images with the application trained models **OR** train the application model with new dataset and test images with this new trained model.
You Can watch Application Demo From [Youtube](https://www.youtube.com/watch?v=8les9jfMM-U&t=111s)
### Models
1. Error Level Analysis"ELA" **[1][2]** top Accuracy **(94.54% , epoc12)** You can read More about ELA from [Here!](https://fotoforensics.com/tutorial-ela.php).
2. VGG16 Pretraind Model.
3. VGG19 Pretraind Model.
### Datasets
Those Models are trained on Many Datasets to Achieve the highest Accuracy
1. MICCF2000 copyMove Dataset :contains 2000 images (1300 authentic-700 tampered ) color images 2048x1536 pixels.
2. CASIAV2 splicing Dataset :contains 12,614 image (7491 authentic -5123 tampered) color images 384x265 pixels.
### Application Description[libraries , Python version , IDE]
Application coded using GUI library PyQt5, tensorflow Keras API , Numpy ,......etc .IDE used - Pycharm community edition and Anaconda Enviroment with python 3.5.4.
### References
**[1]** Agus Gunawan[1], Holy Lovenia[2], Adrian Hartarto Pramudita[3] "Detection og Image tampering With ELA and Deep learning" Informatics Engineering School of Electrical and Informatics Engineering, Bandung Institute of Technology.
**[2]** Nor Bakiah A. W.[1], Mohd. Yamani I. I. [2], Ainuddin Wahid A. W. [3], Rosli Salleh [4] "An Evaluation of Error Level Analysis in Image Forensics" in IEEE 5th International Conference on System Engineering and Technology, Aug 2015. 10 - 11, UiTM, Shah Alam, Malaysia.
没有合适的资源?快使用搜索试试~ 我知道了~
资源推荐
资源详情
资源评论
收起资源包目录
基于深度学习的图像伪造检测.zip (21个子文件)
基于深度学习的图像伪造检测
Icons
cancel-symbol-transparent-9.png 7KB
repeat-pngrepo-com.png 14KB
icons8-cbs-512.ico 22KB
icons8-cbs-512.png 60KB
create.png 12KB
698827-icon-101-folder-search-512.png 9KB
icons8-faq-100 (1).png 3KB
start.png 9KB
Rocket-icon-blue.png 11KB
698831-icon-105-folder-add-512.png 4KB
Source Code
Main_window_Final.py 3KB
VGG19_Training_Module_Final.py 5KB
Result_Retraind_Window_Final.py 5KB
ELA_Training_Module_Final.py 5KB
help_Window.py 2KB
VGG16_Training_Module_Final.py 4KB
Training_window_Final.py 17KB
Test_window_Final.py 14KB
Test_with_Retraind_Modules.py 16KB
Result_Window_Final.py 4KB
README.md 3KB
共 21 条
- 1
资源评论
- 珍妮哦哦吼2024-04-27这个资源对我启发很大,受益匪浅,学到了很多,谢谢分享~
小码蚁.
- 粉丝: 2520
- 资源: 4057
上传资源 快速赚钱
- 我的内容管理 展开
- 我的资源 快来上传第一个资源
- 我的收益 登录查看自己的收益
- 我的积分 登录查看自己的积分
- 我的C币 登录后查看C币余额
- 我的收藏
- 我的下载
- 下载帮助
安全验证
文档复制为VIP权益,开通VIP直接复制
信息提交成功