Dataset Overview

Dataset Information

• Features:

  • img_id: String type, unique identifier for the image.
  • img_a: Image type, first image.
  • img_b: Image type, second image.
  • img_diff: Image type, difference image.
  • sentences: Sequence of strings, sentences describing the differences.

• Dataset Splits:

  • train: Training set, 9,524 samples, size 1,904,363,199.892 bytes.
  • test: Test set, 1,404 samples, size 268,451,640.804 bytes.
  • val: Validation set, 1,634 samples, size 308,229,248.356 bytes.

• Dataset Size:

  • Download size: 2,292,419,742 bytes
  • Total size: 2,481,044,089.052 bytes

Configuration

• Configuration Name: default
  • Data File Paths:
    • Training: data/train-*
    • Testing: data/test-*
    • Validation: data/val-*

Original Dataset

• Source: https://github.com/harsh19/spot-the-diff/

Reference

@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} }