README.md

metadata

title: MFFT AI Image Detector API
emoji: 🔍
colorFrom: blue
colorTo: green
sdk: docker
app_port: 7860
pinned: false
license: mit

MFFT — Multi-Frequency Fusion Transformer, Detection API

FastAPI inference server for AI-generated image detection. Three model variants are served and selectable per request.

Demo checkpoints: these are pilot models (1 training epoch on a 5K subset) published for pipeline demonstration. Full-scale weights will replace them after the training campaign completes.

Endpoints

• GET / — health
• GET /models — available variants (tiny 372K, base 1.62M, large 6.30M)
• POST /predict?model=base — multipart file=<image>; returns prediction, probabilities, anomaly heatmap (base64 PNG), frequency band contributions
• POST /predict/batch?model=large — multiple files

Example

curl -X POST "https://mohsineli-mfft-api.hf.space/predict?model=large" \
  -H "Authorization: Bearer free" \
  -F "file=@image.jpg"