🔎 OCR Admin
engine…
← Workspaces
Logout
📄 Extract — create OCR job
File (PDF / image)
PDFs with a text layer skip OCR automatically.
…or file URL
Doc type
Languages (comma)
Model key (optional)
Confidence threshold
Document ID (optional)
Run OCR
✏️ Correction — feed the fine-tuning dataset
Job ID
Doc type
Field code
Corrected value
Original value
Page
Crop image (optional — the (image → text) training pair)
Submit correction
🎯 Fine-tuning run
Doc type
Component
recognition
detection
Epochs
Batch
Base model
Start run
🧠 Available models
↻