SROIE 2019 V2: ICDAR 2019 Scanned Receipts OCR Dataset

ICDAR 2019 Robust Reading Challenge dataset for Scanned Receipts OCR and Information Extraction. The dataset was sourced from the competition website and processed by a GitHub contributor to remove duplicate images and annotations. It was last updated on the Hugging Face platform on March 30, 2024.

Use Cases

• Train optical character recognition models based on scanned receipt images.
• Benchmark information extraction systems on receipt data.
• Develop models for structured data extraction from receipts.
• Participate in or reproduce results for the ICDAR 2019 Robust Reading Challenge.

Strengths

• Dataset is derived from a well-known international computer vision competition (ICDAR 2019 Robust Reading Challenge).
• Duplicates were removed by a third-party GitHub repository, likely improving data quality.

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

ICDAR 2019 Robust Reading Challenge (https://rrc.cvc.uab.es/?ch=13&com=downloads)

Collection Method

Duplicates removed by https://github.com/Losyash/SROIE-datasetv2

Freshness

Last updated 2024-03-30 18:47:54; freshness should be verified.