|
|
--- |
|
|
license: mit |
|
|
language: |
|
|
- en |
|
|
metrics: |
|
|
- accuracy |
|
|
library_name: tf-keras |
|
|
pipeline_tag: image-classification |
|
|
--- |
|
|
--- |
|
|
|
|
|
# CNN-Based Deepfake Detection Model |
|
|
|
|
|
This repository contains a Convolutional Neural Network (CNN)-based model fine-tuned for deepfake detection. The model has been trained to classify images as either "real" or "fake" (deepfake) using a custom dataset of processed images. |
|
|
|
|
|
## Model Overview |
|
|
|
|
|
This model is a custom CNN architecture built specifically for deepfake detection. It has been designed to efficiently distinguish between real and fake images through a series of convolutional and pooling layers, followed by fully connected layers for classification. |
|
|
|
|
|
### Key Features: |
|
|
- **Model Architecture:** Convolutional Neural Network (CNN) |
|
|
- **Input Size:** 128x128 pixels |
|
|
- **Number of Classes:** 2 (Real, Fake) |
|
|
- **Activation Function:** ReLU in hidden layers, Sigmoid for binary classification |
|
|
- **Regularization:** L2 regularization and Dropout layers to prevent overfitting |
|
|
- **Optimizer:** Adam with a learning rate of 0.0001 |
|
|
- **Training Epochs:** 100 epochs (with early stopping based on validation loss) |
|
|
|
|
|
## Training Details |
|
|
|
|
|
The model was trained on a custom dataset of real and deepfake images, using data augmentation techniques to improve generalization. The training process involved the following components: |
|
|
|
|
|
- **Data Augmentation:** Random rotations, shifts, flips, and brightness adjustments. |
|
|
- **Loss Function:** Binary Cross-Entropy Loss |
|
|
- **Optimizer:** Adam with a learning rate of 0.0001 |
|
|
- **Callbacks:** Early stopping, learning rate scheduler, and model checkpointing were used to optimize training. |
|
|
|
|
|
## Model Performance |
|
|
|
|
|
The model was evaluated on a held-out test set. Below is the key performance metric: |
|
|
|
|
|
- **Test Accuracy:** 71% |
|
|
|
|
|
This accuracy reflects the model's ability to correctly identify real and deepfake images. |
|
|
|
|
|
## Usage |
|
|
|
|
|
You can use this model for inference by loading the model and running predictions on new images. Below is an example using TensorFlow/Keras: |
|
|
|
|
|
```python |
|
|
from tensorflow.keras.models import load_model |
|
|
from tensorflow.keras.preprocessing import image |
|
|
import numpy as np |
|
|
|
|
|
# Load the trained model |
|
|
model = load_model('cnn_model.h5') |
|
|
|
|
|
# Load and preprocess the image |
|
|
img_path = 'path_to_your_image.jpg' |
|
|
img = image.load_img(img_path, target_size=(128, 128)) |
|
|
img_array = image.img_to_array(img) / 255.0 |
|
|
img_array = np.expand_dims(img_array, axis=0) |
|
|
|
|
|
# Make a prediction |
|
|
prediction = model.predict(img_array) |
|
|
print('Real' if prediction[0][0] < 0.5 else 'Fake') |
|
|
``` |
|
|
|
|
|
## How to Use |
|
|
|
|
|
1. **Clone the repository**: |
|
|
```bash |
|
|
git clone https://huggingface.co/MaanVad3r/DeepFake-Detector |
|
|
cd DeepFake-Detection-model.git |
|
|
``` |
|
|
|
|
|
2. **Run Inference**: |
|
|
Use the provided script or the sample code above to run inference on your images. |
|
|
|
|
|
## License |
|
|
|
|
|
This project is licensed under the MIT License. Feel free to use and modify the model as needed. |
|
|
|
|
|
--- |
