Announcement
Indra introduced as a 1-million-parameter OCR model
The introductory post claims 49 pages per second on a CPU—11 times PaddleOCR's speed and 17 times Tesseract's.
TLDR
The post introducing Indra says the 1-million-parameter OCR model processes 49 pages per second on a CPU, 11 times faster than PaddleOCR and 17 times faster than Tesseract. It also reports an OmniParseBench text score of 41.3, versus 35.8 and 19.3 for those models, respectively, and says Indra provides full layout and line bounding boxes.
Combined views
—
2 Sources, first seen ago
