metadata
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
dataset_info:
features:
- name: image
dtype: image
- name: label
dtype:
class_label:
names:
'0': forged
'1': genuine
splits:
- name: train
num_bytes: 10544187.713
num_examples: 6171
download_size: 2251347922
dataset_size: 10544187.713
license: cc-by-4.0
task_categories:
- image-classification
size_categories:
- 1K<n<10K
COMPTECH2022 "WhoSigned?"
Dataset Description
Description
Dataset is containing over 5000 handwritten signatures with correspondent images and crops for real and forged signatures
Each image contains about 10 handwritten signatures from the same user id. The image is then cropped with help of a segmentation neural network. Every crop contains one handwritten signature. You can define id from the image filename.
Created with COMPTECH2022 support by Toloka.ai