SDXL Detector
| Property | Value |
|---|---|
| Parameter Count | 86.8M |
| License | CC-BY-NC-3.0 |
| Architecture | Swin Transformer |
| Validation Metrics | F1: 0.973, Accuracy: 0.981 |
What is sdxl-detector?
SDXL Detector is a specialized image classification model designed to distinguish between authentic images and those generated by the SDXL AI model. Built upon the umm-maybe AI art detector, this model has been fine-tuned on a carefully curated dataset of Wikimedia-SDXL image pairs, where SDXL-generated images are created using BLIP-generated captions from Wikimedia images.
Implementation Details
The model leverages the Swin Transformer architecture and was trained using AutoTrain technology. It demonstrates exceptional performance metrics, including a 0.973 F1 score and 0.981 accuracy on validation data. The model processes both I64 and F32 tensor types and has been optimized for ONNX runtime compatibility.
- Fine-tuned on Wikimedia-SDXL image pairs
- Improved detection capability for recent diffusion models
- Optimized for non-artistic imagery detection
- Supports inference endpoints for deployment
Core Capabilities
- High-precision detection (0.994 precision score)
- Robust performance on modern diffusion model outputs
- Effective handling of diverse image types
- Enhanced detection for non-artistic imagery
Frequently Asked Questions
Q: What makes this model unique?
This model specifically excels at detecting SDXL-generated images with improved accuracy compared to its predecessor, especially for non-artistic imagery and recent diffusion model outputs. Its training on Wikimedia-based data provides a broader subject range coverage.
Q: What are the recommended use cases?
The model is ideal for non-commercial applications in educational and personal use cases for detecting AI-generated images, particularly those created by SDXL. Due to its licensing restrictions (CC-BY-NC-3.0), it should not be used for commercial purposes.
