Skip to content

Dataset · priyank-m

Receipt text crops for OCR recognition

priyank-m/SROIE_2019_text_recognition

Over 52k cropped text fragments from English scanned receipts for training OCR models.

What's inside

Inside are images of individual text blocks cropped along bounding boxes from 973 scanned receipts of the original SROIE set. The dataset is split into 33,626 training images and 18,704 test images in English. Text markup for all fragments within each split is stored in the metadata.jsonl file.

What you can do with it

  • Training and fine-tuning text recognition (OCR) models on documents and receipts
  • Testing the quality of image-to-text systems on small text fragments
Tasks
Image → text
Languages
English
Size
10K–100K rows
Downloads this month
504
Likes
15

How to load it

from datasets import load_dataset

ds = load_dataset("priyank-m/SROIE_2019_text_recognition")