Datasets:
metadata
language:
- en
size_categories:
- 1K<n<10K
task_categories:
- image-to-text
dataset_info:
- config_name: IC03
features:
- name: image
dtype: image
- name: text
dtype: string
splits:
- name: train
num_examples: 1156
- name: test
num_examples: 1110
- config_name: default
features:
- name: image
dtype: image
- name: text
dtype: string
splits:
- name: train
num_bytes: 33972325.5
num_examples: 1156
- name: test
num_bytes: 39217342.25
num_examples: 1110
download_size: 151135637
dataset_size: 73189667.75
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
- split: test
path: data/test-*
ICDAR2003 (IC03) was developed specifically for the ICDAR 2003 Robust Reading competition, aimed at enhancing the recognition of camera-captured scene texts. The dataset comprises 1,156 images for training and an additional 1,110 images for evaluation.