Image Tampering Recognition using YOLOv5

This repository contains an implementation of image tampering recognition using YOLOv5, based on the paper "Image Tampering Recognition Algorithm using Improved YOLOv5" by Z. Liu.

Paper Information

• Title: Image Tampering Recognition Algorithm using Improved YOLOv5
• Paper Link: IEEE Xplore
• Citation: Z. Liu, "Image Tampering Recognition Algorithm Based on Improved YOLOv5s," in IEEE Access, vol. 11, pp. 95114-95119, 2023, doi: 10.1109/ACCESS.2023.3311474.
• Keywords: Feature extraction, Image recognition, Neck, Prediction algorithms, Data augmentation, Object detection, Ethics, Biomedical image processing, Image tamper recognition, YOLOv5s, attention module, EIOU loss function

Installation

To use this implementation, you need to follow these steps:

1. Clone the YOLOv5 model repository:
2. Install the required packages:
3. Install Roboflow:
4. Install ClearML for visualization:

Dataset

The dataset used for this implementation is obtained from Roboflow Universe. You can find the dataset here.

• Total Images: 7257
  • Train: 5075
  • Valid: 1459
  • Test: 723

Customization

To adapt the YOLOv5 architecture for image tampering recognition, the number of classes has been modified from 80 to 2 based on the dataset.

Usage

1. Training: Use the provided dataset to train the YOLOv5 model. You can train the model using the following command:

2. Evaluation: After training, evaluate the model's performance using the validation dataset:

3. Testing: Finally, test the trained model using the test dataset:

Feel free to explore and modify the code to suit your specific requirements.

License

This project is licensed under the MIT License - see the LICENSE file for details.