High Quality Invoice Images for OCR with Structured Metadata

8,181 high-quality synthetic invoice images for OCR and document understanding tasks, created by Voxel51. The dataset includes 1,489 fully annotated samples with structured JSON metadata and raw OCR text, plus 6,692 unannotated images.

Use Cases

• Train OCR models based on high-quality synthetic invoice images.
• Develop document understanding systems based on structured JSON metadata annotations.
• Perform semi-supervised learning experiments based on the mix of annotated and unannotated samples.
• Create or augment annotation projects for invoice processing based on the unannotated image pool.

Strengths

• Contains 8,181 high-quality synthetic images, providing a substantial base for training.
• Includes 1,489 fully annotated samples with structured JSON metadata and OCR text.
• Offers 6,692 unannotated images for tasks like semi-supervised learning.

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.
• Data may reflect synthetic bias inherent to its generation method.

Provenance

Source

Voxel51 via Hugging Face.

Collection Method

Synthetic generation.

Freshness

Last updated 2026-02-05 18:36:21; freshness should be verified.