metadata
task_categories:
- image-to-text
language:
- en
size_categories:
- 100M<n<1B
dataset_name: uparxive-ultexB
license: mit
features:
token_ids:
feature:
type: sequence
feature:
type: sequence
feature:
dtype: int32
text_types:
feature:
type: sequence
feature:
dtype: int8
bboxes:
feature:
type: sequence
feature:
type: array2d
shape:
- 2
- 2
dtype: float32
image:
feature:
type: image
decode: true
splits:
train:
num_examples: 1000
path: Train/*
test:
num_examples: 200
path: Test/*
valid:
num_examples: 200
path: Valid/*