README.md

---
license: apache-2.0
datasets:
- competitions/aiornot
language:
- en
library_name: transformers
base_model:
- google/siglip2-base-patch16-224
pipeline_tag: image-classification
tags:
- SigLIP2
- AI-vs-Real
- art
---

![1.png](https://cdn-uploads.huggingface.co/production/uploads/65bb837dbfb878f46c77de4c/KhTM3MJls_zbF4q2EqxUO.png)

# AIorNot-SigLIP2

> AIorNot-SigLIP2 is a vision-language encoder model fine-tuned from google/siglip2-base-patch16-224 for binary image classification. It is trained to detect whether an image is generated by AI or is a real photograph using the SiglipForImageClassification architecture.

> [!note]
*SigLIP 2: Multilingual Vision-Language Encoders with Improved Semantic Understanding, Localization, and Dense Features* https://arxiv.org/pdf/2502.14786


```py
Classification Report:
              precision    recall  f1-score   support

        Real     0.9215    0.8842    0.9025      8288
          AI     0.9100    0.9396    0.9246     10330

    accuracy                         0.9149     18618
   macro avg     0.9158    0.9119    0.9135     18618
weighted avg     0.9151    0.9149    0.9147     18618
```

![download (2).png](https://cdn-uploads.huggingface.co/production/uploads/65bb837dbfb878f46c77de4c/AtUeLO-22whnmrN-mJZDr.png)

---

## Label Space: 2 Classes

The model classifies an image as either:

```
Class 0: Real
Class 1: AI
```

---

## Install Dependencies

```bash
pip install -q transformers torch pillow gradio hf_xet
```

---

## Inference Code

```python
import gradio as gr
from transformers import AutoImageProcessor, SiglipForImageClassification
from PIL import Image
import torch

# Load model and processor
model_name = "prithivMLmods/AIorNot-SigLIP2"  # Replace with your model path
model = SiglipForImageClassification.from_pretrained(model_name)
processor = AutoImageProcessor.from_pretrained(model_name)

# Label mapping
id2label = {
    "0": "Real",
    "1": "AI"
}

def classify_image(image):
    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()

    prediction = {
        id2label[str(i)]: round(probs[i], 3) for i in range(len(probs))
    }

    return prediction

# Gradio Interface
iface = gr.Interface(
    fn=classify_image,
    inputs=gr.Image(type="numpy"),
    outputs=gr.Label(num_top_classes=2, label="AI or Real Detection"),
    title="AIorNot-SigLIP2",
    description="Upload an image to classify whether it is AI-generated or Real."
)

if __name__ == "__main__":
    iface.launch()
```

---

## Intended Use

AIorNot-SigLIP2 is useful in scenarios such as:

* AI Content Detection – Identify AI-generated images for social platforms or media verification.
* Digital Media Forensics – Assist in distinguishing synthetic from real-world imagery.
* Dataset Filtering – Clean datasets by separating real photographs from AI-synthesized ones.
* Research & Development – Benchmark performance of image authenticity detectors.