ChineseOCRBench: A Benchmark for Evaluating Chinese OCR in Multimodal Models

ChineseOCRBench is a dataset created to evaluate the performance of large multimodal models on Chinese Optical Character Recognition tasks. The dataset was extracted from the work 'On the Hidden Mystery of OCR in Large Multimodal Models' and published by author SWHL on Hugging Face in April 2024. It serves as a dedicated benchmark for Chinese OCR evaluation.

Use Cases

• Benchmarking OCR accuracy of multimodal models based on Chinese text images.
• Training OCR models specifically for Chinese language text recognition.
• Comparing model performance on a dedicated Chinese OCR evaluation task.

Strengths

• Dataset is specifically designed to address a gap in Chinese OCR evaluation for multimodal models.
• It is derived from a published research work, providing a foundation for comparative analysis.

Limitations

• Description metadata is limited; actual data quality requires manual inspection after download.
• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.

Provenance

Source

SWHL on Hugging Face

Collection Method

Extracted from the research paper 'On the Hidden Mystery of OCR in Large Multimodal Models'.

Freshness

Last updated 2024-04-30 09:40:36; freshness should be verified.