Ocular OCR Service

Extract text and data from PDFs and images using advanced OCR technology

Unified OCR Workflow

Select providers, configure strategies, and process multiple documents in one place

Upload Files
Drop files here or click to browse

Supported formats: PDF, JPG, PNG, BMP, TIFF, WEBP

Processing Options
Choose how the platform coordinates multiple OCR providers.
OCR Providers
No OCR providers available. Check your configuration.
Select at least one file to continue.
Processing Strategies
Single: Use one provider
Fallback: Try providers until success
Ensemble: Use multiple providers
Best: Choose best result by confidence
Loading...

Processing Files...

Please wait while we extract text from your documents

OLMocr Document Processor

Convert PDFs and images to markdown using AI-powered OCR technology

Supported formats: PDF, PNG, JPG, JPEG (max 50MB)
Enter a direct link to a PDF or image file
AI-Powered OCR

Uses OLMocr for accurate document conversion

Structure Preservation

Maintains document structure, tables, equations, and reading order

Markdown Output

Clean, structured markdown format ready for publishing

Real Estates Auction Valuation Template

Complete E2E processing: OCR extraction + structured data extraction + due diligence report generation

Upload Property Auction Documents
Drop PDF files here or click to browse

Upload multiple property auction PDFs (appraisals, due diligence reports, cadastral documents)

OCR engine for text extraction from PDFs
Strategy for OCR processing
Language hint for better OCR accuracy
Number of pages to process in parallel

UTF-8 tidying, mojibake repair, safe corrections
Fixes common digit OCR errors (e.g., O→0, l→1)
Reconstruct paragraphs from page-based text
Repair encoding errors (e.g., "Ã " → "à")

Select LLM for extraction (multi-select enabled for Ensemble strategy)
Schema defines what fields to extract Use 0.0 for consistent extraction
Optional custom prompt for LLM extraction

Store extraction results for later retrieval
Results will be available in JSON format after processing completes.
Process Summary
OCR: OLM OCR → Corrections: All → LLM: Mistral 7B (1 model) → Schema: PropertyAppraisalDocument → Save: Yes
Workflow Pipeline
1
OCR Extraction
Provider: OLM OCR
Strategy: Single
Workers: 4
2
Text Corrections
Mode: All Corrections
Digit Fix: Auto
Reflow: Yes, Mojibake: Yes
3
LLM Extraction
LLM(s): Mistral 7B
Count: 1 model
Schema: PropertyAppraisalDocument
Temp: 0.0
4
Output
Save to DB: Yes
JSON format
Expected Output
Sale ID & Tribunal info
Lot numbers & asset details
Cadastral data (Foglio, Particella, etc.)
Valuations & base prices
Liens & encumbrances
Compliance issues
Structured JSON export

Property Extraction with OLMocr Pipeline

OCR extraction using OLMocr → Text corrections → LLM structure extraction (simplified workflow)

Upload Property Documents
Drop PDF files here or click to browse

Upload property auction PDFs for OLMocr processing

Using OLMocr v2 (allenai/olmocr-2-7b-1025) via DeepInfra cloud GPU API
OLMocr v2 model variant
Cloud GPU service provider
Optional: custom endpoint or deployment ID
Concurrent API calls to DeepInfra

LLM extraction runs on cloud GPU services (DeepInfra for open models, API calls for proprietary)
Ctrl+Click to multi-select. Shows combined models from selected providers.
Ctrl+Click to multi-select for ensemble extraction
User prompt template for LLM extraction (schema and document text will be auto-inserted)
Loaded via /api/schema/property-appraisal
The backend automatically loads the full PropertyAppraisalDocument schema for structured extraction
Pipeline Steps
1
OCR Extraction (OLMocr)
↓
2
Text Corrections
↓
3
LLM Extraction
1 model
↓
4
JSON Output
Expected Fields
Sale ID & Tribunal
Lot numbers
Cadastral data
Valuations
Liens & encumbrances
Structured JSON