OpenSDI-SD1.5-SigLIP2

OpenSDI-SD1.5-SigLIP2

OpenSDI-SD1.5-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 Stable Diffusion 1.5 (SD1.5), utilizing 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 SD1.5 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 the OpenSDI-Flux.1-SigLIP2 model.

Classification Report:
                 precision    recall  f1-score   support

     Real_Image     0.9036    0.9323    0.9177     10000
SD1.5_Generated     0.9301    0.9005    0.9150     10000

       accuracy                         0.9164     20000
      macro avg     0.9168    0.9164    0.9164     20000
   weighted avg     0.9168    0.9164    0.9164     20000

---

Label Space: 2 Classes

The model classifies an image as either:

Class 0: Real_Image
Class 1: SD1.5_Generated

---

Install Dependencies

pip install -q transformers torch pillow gradio hf_xet

---

Inference Code

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

# Load model and processor
model_name = "prithivMLmods/OpenSDI-SD1.5-SigLIP2"  # Replace with your model path
model = SiglipForImageClassification.from_pretrained(model_name)
processor = AutoImageProcessor.from_pretrained(model_name)

# Label mapping
id2label = {
    "0": "Real_Image",
    "1": "SD1.5_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="SD1.5 Image Detection"),
    title="OpenSDI-SD1.5-SigLIP2",
    description="Upload an image to determine whether it is a real photograph or generated by Stable Diffusion 1.5 (SD1.5)."
)

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

---

Intended Use

OpenSDI-SD1.5-SigLIP2 is designed for the following use cases:

• Generative Model Evaluation – Detect SD1.5-generated images for analysis and benchmarking.
• Dataset Integrity – Filter out AI-generated images from real-world image datasets.
• Digital Media Forensics – Support visual content verification and source validation.
• Trust & Safety – Detect synthetic media used in deceptive or misleading contexts.