cff-version: 1.2.0
title: SimdPaddleOCR Synthetic OCR Benchmark v1
message: If you use this dataset, please cite it and the SimdPaddleOCR project.
type: dataset
authors:
  - name: Sdcb
    alias: sdcb
license: Apache-2.0
repository-code: https://github.com/sdcb/SimdPaddleOCR
url: https://huggingface.co/datasets/sdcb/simdpaddleocr-dataset-v1
abstract: >
  A compact, deterministic, bilingual (zh/en) synthetic OCR benchmark of 100
  JPEG images and 1,036 line-level annotations. Generated with a fixed seed
  for SimdPaddleOCR performance and accuracy regression.
keywords:
  - ocr
  - synthetic
  - benchmark
  - paddleocr
  - chinese
  - english
version: "1.0.0"
date-released: "2026-09-16"