⬢github Python · 950 ★ +63 since we first saw it · pushed 18 h ago · Apache-2.0
datalab-to/indra
1M param OCR model - runs on CPU, 15x faster than tesseract, close to 1B+ param models in accuracy
Indra OCR is a tiny 1M-parameter OCR model from Datalab that runs on CPU via ONNX Runtime. It converts PDFs and images to markdown, text, or JSON with layout boxes, supporting handwriting, 80+ languages, rotated scans, and digital PDF text layers. Benchmarks claim it's 15x faster than Tesseract with accuracy ahead of PaddleOCR.
Why now: It's trending as a newly created repo appearing among the most-starred GitHub repositories, likely driven by its unusually good speed/accuracy tradeoff for a 1M-param CPU-only model.
Who it is for: Developers and teams needing fast, cheap OCR and document parsing at scale without GPUs.
aicpudocument-parsingocronnxpdf
Stars over our 81 snapshots: 887 to 950, since 1 d ago.
Where people talked about it
- ⬢github new repos, most starred 12 min ago
- ⬢github new repos, most starred 5 h ago
API: https://socialmediatrends-api.osmike.com/v1/repos/datalab-to/indra