Deepfake-vs-Real-8000 is an image classification vision-language encoder model fine-tuned from google/siglip2-base-patch16-224 for a single-label classification task. It is designed to detect whether an image is a deepfake or a real one using the SiglipForImageClassification architecture.

Classification Report:
              precision    recall  f1-score   support

    Deepfake     0.9990    0.9972    0.9981      4000
    Real one     0.9973    0.9990    0.9981      4000

    accuracy                         0.9981      8000
   macro avg     0.9981    0.9981    0.9981      8000
weighted avg     0.9981    0.9981    0.9981      8000

The model categorizes images into two classes:

Class 0: "Deepfake"
Class 1: "Real one"

Run with Transformers🤗

!pip install -q transformers torch pillow gradio

python

import gradio as gr
from transformers import AutoImageProcessor
from transformers import SiglipForImageClassification
from transformers.image_utils import load_image
from PIL import Image
import torch

# Load model and processor
model_name = "prithivMLmods/Deepfake-vs-Real-8000"
model = SiglipForImageClassification.from_pretrained(model_name)
processor = AutoImageProcessor.from_pretrained(model_name)

def deepfake_classification(image):
    """Predicts whether an image is a Deepfake or Real."""
    image = Image.fromarray(image).convert("RGB")
    inputs = processor(images=image, return_tensors="pt")
    
    with torch.no_grad():
        outputs = model(**inputs)
        logits = outputs.logits
        probs = torch.nn.functional.softmax(logits, dim=1).squeeze().tolist()
    
    labels = {
        "0": "Deepfake", "1": "Real one"
    }
    predictions = {labels[str(i)]: round(probs[i], 3) for i in range(len(probs))}
    
    return predictions

# Create Gradio interface
iface = gr.Interface(
    fn=deepfake_classification,
    inputs=gr.Image(type="numpy"),
    outputs=gr.Label(label="Prediction Scores"),
    title="Deepfake vs. Real Image Classification",
    description="Upload an image to determine if it's a Deepfake or a Real one."
)

# Launch the app
if __name__ == "__main__":
    iface.launch()

Intended Use:

The Deepfake-vs-Real-8000 model is designed to detect deepfake images from real ones. Potential use cases include:

Deepfake Detection: Assisting cybersecurity experts and forensic teams in detecting synthetic media.
Media Verification: Helping journalists and fact-checkers verify the authenticity of images.
AI Ethics & Research: Contributing to studies on AI-generated content detection.
Social Media Moderation: Enhancing tools to prevent misinformation and digital deception.