| --- |
| 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 |
| - AI-vs-Real |
| - SigLIP2 |
| - SD2.1 |
| --- |
| |
|  |
| |
| # OpenSDI-SD2.1-SigLIP2 |
|
|
| > OpenSDI-SD2.1-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 Stable Diffusion 2.1 (SD2.1), 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 |
|
|
| > [!note] |
| *OpenSDI: Spotting Diffusion-Generated Images in the Open World* https://arxiv.org/pdf/2503.19653, OpenSDI SD2.1 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 the OpenSDI-Flux.1-SigLIP2 model. |
|
|
| ```py |
| Classification Report: |
| precision recall f1-score support |
| |
| Real_Image 0.8551 0.8967 0.8754 10000 |
| SD2.1_Generated 0.8914 0.8481 0.8692 10000 |
| |
| accuracy 0.8724 20000 |
| macro avg 0.8733 0.8724 0.8723 20000 |
| weighted avg 0.8733 0.8724 0.8723 20000 |
| ``` |
|
|
|  |
|
|
| --- |
|
|
| ## Label Space: 2 Classes |
|
|
| The model classifies an image as either: |
|
|
| ``` |
| Class 0: Real_Image |
| Class 1: SD2.1_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-SD2.1-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": "SD2.1_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="SD2.1 Image Detection"), |
| title="OpenSDI-SD2.1-SigLIP2", |
| description="Upload an image to determine whether it is a real photograph or generated by Stable Diffusion 2.1 (SD2.1)." |
| ) |
| |
| if __name__ == "__main__": |
| iface.launch() |
| ``` |
|
|
| --- |
|
|
| ## Intended Use |
|
|
| OpenSDI-SD2.1-SigLIP2 is designed for the following applications: |
|
|
| * Generative Image Detection – Identify SD2.1-generated images for auditing or validation. |
| * Dataset Curation – Clean datasets by removing synthetic images. |
| * Visual Authenticity Verification – Distinguish real images from AI-generated ones. |
| * Digital Forensics – Assist in tracing the source of digital images in investigative workflows. |
