--- title: README emoji: 🔍 colorFrom: blue colorTo: indigo sdk: static pinned: false --- # Remtir **Remtir is a Thai AI research lab working on computer vision.** We focus on what Thai documents and images need, which global models often get wrong: stacked vowels and tone marks, Thai digits, government forms, dense tables and pages photographed on a phone. ห้องวิจัย AI ไทยด้าน Computer Vision พัฒนาโมเดลอ่านเอกสารและภาพภาษาไทยแบบเปิด ## What we work on - **Document AI:** OCR, layout analysis, reading order and table structure for Thai and English documents - **Text recognition:** recognizers built around Thai script, such as grapheme-cluster CTC - **Efficient vision models:** small models that run on CPU and on-premise, so documents never have to leave the building - **Evaluation:** fair, reproducible benchmarks for Thai OCR ## Projects | Project | What it is | |---|---| | [AksonOCR-Lite](https://github.com/Remtir/AksonOCR-Lite) | Open-source Thai document OCR on ONNX Runtime: PDF/image → Markdown, layout and tables, ~1.5 s/page on CPU. Apache-2.0. [Live demo](https://aksonocr.com/lite-demo) · [Weights](https://huggingface.co/Remtir/AksonOCR-Lite) | | [AksonOCR](https://aksonocr.com) | Our document parsing API for harder pages: handwriting, charts, formulas and key extraction. |