capcheck/ai-image-detection

Detects if an image is real or fake.

Public
67 runs
License
Weights
Playground API Examples README Versions

Run time and cost

This model runs on Nvidia T4 GPU hardware. We don't yet have enough runs of this model to provide performance information.

Readme

CapCheck AI Image Detection

Detect AI-generated images using a Vision Transformer (ViT) model. Returns confidence scores for real vs AI-generated classification.

Use Cases

  • Fact-checking: Verify if images in news articles are AI-generated
  • Content moderation: Flag synthetic media before publication
  • Research: Analyze datasets for AI-generated content

API

Input

Parameter Type Default Description
image file/URL required Image to analyze (JPEG, PNG, WebP)
threshold float 0.5 Classification threshold (0.0-1.0)

Output

{
  "is_ai_generated": true,
  "confidence": 0.97,
  "ai_probability": 0.97,
  "real_probability": 0.03,
  "threshold": 0.5,
  "model_version": "v1.0.0",
  "inference_time_ms": 45.2
}

Model Info

Property Value
Model capcheck/ai-image-detection
Architecture ViT-Base (86M parameters)
Input 224x224 pixels
Output Real/Fake classification
License Apache 2.0

Model Lineage

This model builds on excellent open-source work:

  1. Google - ViT-Base architecture (Apache 2.0)
  2. dima806 - Fine-tuned on CIFAKE dataset for AI detection
  3. CapCheck - Published with ongoing improvements for modern AI generators
Model created
Model updated