Abstract
A Thai OCR model trained solely on synthetic documents achieves strong real-world performance by isolating key transfer factors like typography, layout, and handwriting glyphs.
We investigate what makes synthetic OCR supervision transfer to real Thai documents and use the resulting insights to build Wayu-Paxa-OCR-Zero, a Thai OCR model adapted without OCR labels from real Thai document pages. Synthetic data provide exact labels at scale, but "realism" conflates source domain, page context, typography, spatial structure, and glyph variation. We disentangle these factors with a controlled document-reconstruction pipeline and evaluate each variant under page- and crop-level training on printed and handwritten Thai documents. Non-text context has little consistent effect, whereas typeface diversity, two-dimensional structure, and real handwriting glyphs improve transfer; moreover, source-domain matching depends on training granularity, with in-domain reconstruction approaching real printed supervision under page-level training (1.82% versus 1.31% median character error rate) but underperforming out-of-domain reconstruction under crop-level training (15.59% versus 5.52%). Guided by these findings, we adapt the 0.9B-parameter PaddleOCR-VL-1.6 into Wayu-Paxa-OCR-Zero using 45,723 synthetic pages: relative to its base checkpoint, it reduces median character error rate from 6.64% to 1.24% on printed pages and from 74.87% to 20.55% on handwriting and outperforms Typhoon OCR v1 7B on all five evaluation sets, showing that synthetic-only training can be competitive.
Community
Hi
Kunat Pipatanakul,
Congratulations on this project. I read your paper and found it both practical and very interesting. It is clear that a great deal of effort went into preparing the training data and developing the model.
I’m currently working part-time as an AI Engineer in Bangkok. At my company, we frequently translate Thai PDF documents into English using frontier models. I have also experimented with building OCR solutions, but I have found it challenging to achieve both fast and reliable performance for Thai documents.
After reading the model card for your 0.9B-parameter model, I’m very excited to try it in our workflow. I would love to connect, share any results or progress from testing it, and potentially contribute if there is an opportunity.
Thank you again for making this work available!
This is an automated message from the Librarian Bot. I found the following papers similar to this paper.
The following papers were recommended by the Semantic Scholar API
- When Do VLMs Help Arabic Manuscript OCR? A Cross-Dataset Study (2026)
- Combining Synthetic and Real Data for Low-Resource Historical OCR: A Manchu Case Study (2026)
- OmniHandwritingOCR: A Diagnostic Benchmark for Evaluating Multimodal LLMs in Handwritten OCR Scenarios (2026)
- Synth-JDoc: Synthesizing a Japanese Document Image Dataset for OCR with Diverse Layouts and Embedded Images (2026)
- Impact of Iterative Fine-Tuning on Transcription Accuracy in Complex Historical Sanskrit Manuscripts (2026)
- TongGuOCR: A Layout-Aware and Token-Augmented OCR MLLM for Chinese Historical Documents (2026)
- Learning to Ground Before Reading: Unified PCB Engineering Drawing Parsing with Compact Vision-Language Models (2026)
Please give a thumbs up to this comment if you found it helpful!
If you want recommendations for any Paper on Hugging Face checkout this Space
You can directly ask Librarian Bot for paper recommendations by tagging it in a comment: @librarian-bot recommend
Get this paper in your agent:
hf papers read 2609.03595 Don't have the latest CLI?
curl -LsSf https://hf.co/cli/install.sh | bash Models citing this paper 1
Datasets citing this paper 0
No dataset linking this paper
Spaces citing this paper 2
Collections including this paper 0
No Collection including this paper