metadata
dataset_info:
features:
- name: img_id
dtype: string
- name: img_a
dtype: image
- name: img_b
dtype: image
- name: img_diff
dtype: image
- name: sentences
sequence: string
splits:
- name: train
num_bytes: 1904363199.892
num_examples: 9524
- name: test
num_bytes: 268451640.804
num_examples: 1404
- name: val
num_bytes: 308229248.356
num_examples: 1634
download_size: 2292419742
dataset_size: 2481044089.052
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
- split: test
path: data/test-*
- split: val
path: data/val-*
Original dataset: https://github.com/harsh19/spot-the-diff/
@inproceedings{jhamtani2018learning,
title={Learning to Describe Differences Between Pairs of Similar Images},
author={Jhamtani, Harsh and Berg-Kirkpatrick, Taylor},
booktitle={Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing (EMNLP)},
year={2018}
}