Extract Text (EasyOCR)

@jaided1 VersionExtract

Detects and recognises text in each Image with EasyOCR and emits either a joined String transcript or per-line regions carrying Polygon<Double> or Rectangle<Double> bounds with optional confidence. Supports multi-language reading.

How it fits

Typical backends

OCR scene images from files and forward the transcript and geometry over HTTP.

Crop a detected region first, then OCR the crop and summarise it with a text model.

Caveats

I/O contractThe output arm is fixed when the backend deploys: a text consumer pins the String transcript, a region consumer pins per-line records with Polygon<Double> or Rectangle<Double> bounds. The polygon carries EasyOCR's native quad and the rectangle its axis-aligned bound.
Parameter interactionlanguage accepts one or several codes separated by a comma or plus sign; common ISO three-letter codes are mapped to two-letter equivalents. The chosen language set must match the script in the images or recognition degrades.
Hard constraintThe CRAFT detector and the recognition network are mounted as file links and loaded locally; nothing is downloaded at runtime. The mounted recognition weights must cover the configured language, or recognition degrades.
Fallbackdevice requesting CUDA resolves to CPU when no compatible accelerator is visible at startup; inference continues at CPU latency.
LimitationEmpty or whitespace-only lines are dropped before emission, so the region arms never contain blank entries.
State lifetimeThe reader is built once per language at startup; changing the language set or models requires a redeploy.

Versionen

f403974elatestdefault
Automated release
25.6.2026