Upload Document
Supports PDF, PNG, JPEG, and WebP (Up to 10MB)
Click to browse or drag and drop
PDF, PNG, JPG, or WebP
Pipeline Architecture
1
Stage 1 (T4 GPU): PP-DocLayoutV3 performs layout detection, identifies tables, formulas, charts, and reading flow.
2
Stage 2 (A10G GPU): Qwen 3.5 transcribes crops into clean Markdown tables and LaTeX formulas.
Upload a document on the left to run two-stage OCR.
Processing on Modal GPU...
Executing PP-DocLayoutV3 & Qwen 3.5 in cloud...