// IMAGE & VISION
OCR and Drawing Text Extraction
Azakros has two text-reading nodes. IMAGE__OCR reads general text in many languages using PaddleOCR. CAD__OCR_DRAWING is aimed at engineering drawings and pulls out dimensions and labels.
Nodes used
IMAGE__OCRText extraction with multi-language support via PaddleOCR.
CAD__OCR_DRAWINGExtracts dimensions and labels from engineering blueprints.
IMAGE__CAPTIONOptional. Describes the image in a sentence for indexing and accessibility.
How to set it up
- 01Add INPUT__IMAGE and connect it to IMAGE__OCR (documents, signs, screenshots) or CAD__OCR_DRAWING (drawings).
- 02Add OUTPUT__ and publish the pipeline.
- 03Send an image_url in the execute request. The recognised text comes back in outputs.
Check the result before you rely on it
OCR on handwriting, skewed scans or tiny annotations can misread characters, and a misread dimension is worse than a missing one. For drawings, run IMAGE__UPSCALE first and review extracted values before they feed manufacturing steps.
Feeding the text onward
The output is plain text, so it can go into an LLM__CHAT node to summarise or classify a document, or out to your own system through the API response or a callback URL.