OCR-Ops is infrastructure to perform Optimal Character Recognition (OCR) at scale on a large number of images and videos. Built on top of the algo-ops framework, OCR-Ops is modular and extensible in its data processing operations.
Key Features:
- Supports building an OCRPipeline that can utilize multiple popular OCR annotation methods (e.g. PyTesseract, EasyOCR, etc.) and return the results in structured and efficient fashion within a unified framework.
- Enables multi-levels of information of the OCR application (e.g. text-only, bounding boxes, etc.)
- Allows definition of an image pre-processing pipeline (before OCR) and a text-cleaning pipeline (after OCR) of detected but noisy text to enable optimal and robust OCR performance.
- Supports several nice presets that are plug-and-play for the above purpose!