OCR Text Detection and Recognition Dataset Aggregating 14 Public Benchmarks

A large-scale OCR dataset aggregates 14 public benchmarks for text detection and recognition. It pairs images with transcribed text, bounding boxes, and polygon coordinates for text regions. The dataset is authored by Yesianrohn and was last updated on HuggingFace in April 2026.

Use Cases

• Train scene text detection models based on bounding box and polygon coordinate annotations.
• Benchmark optical character recognition (OCR) systems based on transcribed text paired with images.
• Develop multi-source OCR pipelines based on the aggregated data from 14 public benchmarks.
• Fine-tune text recognition models for handwritten documents based on the described annotations.

Strengths

• Aggregates data from 14 distinct public benchmarks, suggesting broad coverage.
• Provides multiple annotation types per image: transcribed text, bounding boxes, and polygon coordinates.
• Described as a large-scale dataset, implying a substantial number of samples.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count and total dataset size are unknown, which may limit suitability assessment.
• Freshness should be verified as the last update timestamp is in the future (2026-04-12).

Provenance

Source

Aggregated from 14 public OCR benchmarks.

Collection Method

Aggregation of existing public datasets.

Freshness

Last updated 2026-04-12 12:19:51