Bank Name Region Detection Dataset: 5,976 Annotated Financial Document Images

5,976 images of financial documents annotated for bank name region localization in the COCO object detection format. The dataset is split into 5,580 training, 264 validation, and 132 test images. It was created by Srijan-Chakraborty and last updated on July 1, 2026.

Use Cases

• Training object detection models to locate bank name regions based on annotated bounding boxes.
• Fine-tuning pre-trained vision models for financial document information extraction.
• Benchmarking model performance on a specialized task of bank name region detection.
• Developing OCR-aware models that combine visual detection with text recognition for financial documents.

Strengths

• 5,976 total annotated images provide a substantial base for model training.
• Dataset includes a defined split of 5,580 training, 264 validation, and 132 test images for structured evaluation.

Limitations

• Column-level documentation and specific class details are absent, requiring inspection after download.
• The dataset's source, collection method, and potential geographic or institutional bias are not described.

Provenance

Source

huggingface

Freshness

Last updated 2026-07-01 11:04:59.