ICDAR2019 SROIE: Scanned Receipts for OCR and Information Extraction

ICDAR2019's Scanned Receipts OCR and Information Extraction (SROIE) dataset was originally published by Huang et al. for the 15th International Conference on Document Analysis and Recognition Robust Reading Challenge. This work presents an extension of the original dataset, including 14 receipt annotations missing from the original Task 3 test dataset. The dataset was last updated on the platform in 2026.

Use Cases

• Train optical character recognition models based on scanned receipt images.
• Develop information extraction systems to locate and parse key fields like date, address, and total from receipts.
• Benchmark model performance for the ICDAR Robust Reading Challenge on scanned receipts.
• Fine-tune multimodal models for joint text detection and entity recognition in document images.

Strengths

• Dataset originates from the ICDAR2019 Robust Reading Challenge, a recognized benchmark in document analysis.
• Extension includes 14 previously missing receipt annotations for the Task 3 test set.
• Last updated on the platform on 2026-02-02 00:19:29.

Limitations

• Description metadata is limited; actual data quality requires manual inspection after download.
• Row count, file formats, and column-level documentation are absent.
• License information is unknown, which may restrict usage.

Provenance

Source

Originally published by Huang et al. for the ICDAR2019 Robust Reading Challenge.

Collection Method

Scanned receipts, likely collected and annotated for a competition.

Freshness

Last updated 2026-02-02 00:19:29; freshness should be verified.