---
license: apache-2.0
datasets:
- nebula/OpenSDI_test
- madebyollin/megalith-10m
language:
- en
base_model:
- google/siglip2-base-patch16-224
pipeline_tag: image-classification
library_name: transformers
tags:
- OpenSDI
- Spotting Diffusion-Generated Images in the Open World
- OpenSDI
- Flux.1
- AI-vs-Real
- SigLIP2
---

![1.png](https://cdn-uploads.huggingface.co/production/uploads/65bb837dbfb878f46c77de4c/cYPu51BLrWBZu3GLCmWOe.png)

# OpenSDI-Flux.1-SigLIP2

> OpenSDI-Flux.1-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 a real photograph or generated using the Flux.1 generative model, based on the SiglipForImageClassification architecture.

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

> [!note]
*OpenSDI: Spotting Diffusion-Generated Images in the Open World* https://arxiv.org/pdf/2503.19653, OpenSDI Flux.1 SigLIP2 works best with crisp and high-quality images. Noisy images are not recommended for validation.

> [!warning]
If the task is based on image content moderation or AI-generated image vs. real image classification, it is recommended to use this model.

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

      Real_Image     0.9108    0.9238    0.9172     10000
Flux.1_Generated     0.9227    0.9095    0.9160     10000

        accuracy                         0.9166     20000
       macro avg     0.9167    0.9166    0.9166     20000
    weighted avg     0.9167    0.9166    0.9166     20000
```

![download.png](https://cdn-uploads.huggingface.co/production/uploads/65bb837dbfb878f46c77de4c/M7-dIp4buiR8h6rj5n6Yk.png)

---

## Label Space: 2 Classes

The model classifies an image as either:

```
Class 0: Real_Image
Class 1: Flux.1_Generated
```

---

## 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/OpenSDI-Flux.1-SigLIP2"  # Update if needed
model = SiglipForImageClassification.from_pretrained(model_name)
processor = AutoImageProcessor.from_pretrained(model_name)

# Label mapping
id2label = {
    "0": "Real_Image",
    "1": "Flux.1_Generated"
}

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="Flux.1 Image Detection"),
    title="OpenSDI-Flux.1-SigLIP2",
    description="Upload an image to determine whether it is a real photograph or generated by Flux.1."
)

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

---

## Demo Inference

> [!warning]
Flux.1 Generated

| Image 1 | Image 2 |
|-----|--------|
| ![Mak](https://cdn-uploads.huggingface.co/production/uploads/65bb837dbfb878f46c77de4c/-AGOBIAa4h5Ia5s7l3lka.png) | ![Ethius](https://cdn-uploads.huggingface.co/production/uploads/65bb837dbfb878f46c77de4c/UShWeGHLojpX4P4xEnHSk.png) |

> [!warning]
Real Image

| Image 1 | Image 2 |
|-----|-----|
| ![Image1](https://cdn-uploads.huggingface.co/production/uploads/65bb837dbfb878f46c77de4c/uY2dSKa__zPpoy7wMR9-I.png) | ![Image2](https://cdn-uploads.huggingface.co/production/uploads/65bb837dbfb878f46c77de4c/qoSzI-9ZqsJmg3ltEcxBx.png) |

---

## Intended Use

OpenSDI-Flux.1-SigLIP2 is designed for tasks such as:

* Generative Model Evaluation – Distinguish Flux.1-generated images from real photos for benchmarking and validation.
* Dataset Auditing – Detect synthetic images in real-world datasets to maintain integrity.
* Misinformation Detection – Identify AI-generated visuals in online or news content.
* Media Authentication – Verify whether visual content originates from human-captured or model-generated sources.