proomt

Search

Search posts, papers, and topics

All posts

Hugging Face Daily PapersKunat Pipatanakul1 min readpaperadvanced

How Far Can Synthetic Data Take Thai OCR?

Read on arxiv.org

Summary

The paper dissects which aspects of synthetic Thai OCR data (typeface diversity, 2‑D layout, real handwriting glyphs) actually help transfer to real documents. Using those insights they train Wayu-Paxa-OCR-Zero without any real OCR labels and achieve character error rates comparable to or better than supervised baselines.

  • Typeface diversity, 2‑D page structure, and real handwriting glyphs improve synthetic‑to‑real transfer; non‑text context has little effect.
  • Page‑level training benefits from in‑domain synthetic reconstruction, reaching 1.82% CER vs 1.31% with real printed data.
  • Crop‑level training favors out‑of‑domain synthetic data; in‑domain reconstruction underperforms (15.59% vs 5.52% CER).
  • Wayu-Paxa-OCR-Zero (0.9B model) trained on 45k synthetic pages cuts printed‑page CER from 6.64% to 1.24% and handwriting CER from 74.87% to 20.55%.

OCR engineers targeting low‑resource scripts and teams exploring synthetic data pipelines should see which synthetic factors actually boost real‑world performance.

7/10
computer visionocrsynthetic datanlpthai

Related reading

  1. Building and Evaluating Fixed-Voice Thai TTS from Synthetic Speech

    Researchers train a compact 82 M‑parameter Thai fixed‑voice TTS model using synthetic speech generated by a large voice‑cloning teacher, requiring only a 15‑second real reference. The student achieves 68.2% keyword accuracy and 91.4% pause precision, outperforming its teacher on pause placement and enabling on‑device inference.

    Hugging Face Daily Papersarxiv.org1 minpaper

  4. FLAT: Resampling Image and Text into 1D Flexible-Length Aligned Transmodal Tokens for Retrieval and Generation

    FLAT is a pre‑training framework that encodes images and text into a shared 1D token sequence with variable length via nested dropout, enabling the same embeddings for cross‑modal retrieval and generation. It attains state‑of‑the‑art scores on COCO and Flickr30K for captioning, retrieval, and text‑to‑image generation, and supports interpolation and zero‑shot composed retrieval.

    Hugging Face Daily Papersarxiv.org1 minpaper