Skip to content

clp-research/image-description-sequences

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

27 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

"tell me more": a corpus of image description sequences

We present a dataset of description sequences, a sequence of expressions that together are meant to single out one image from an (imagined) set of other similar images. These sequences were produced in a monological setting, but with the instruction to imagine they were provided to a partner who successively asked for more information (hence, tell me more). The example of such sequence is given below:

sequence

Structure of the repository

├── README.md
├── data
│   ├── sequences.csv
│   └── splits.json
├── papers
│   └── inlg19_short.pdf
└── scripts
    └── download_images.sh

sequences.csv contains the dataset of description sequences. It has the following columns:

  • seq_id: unique id of the sequence
  • image_id: id of the image
  • image_path: path to the image, starting from the directory with all ADE20k images
  • image_cat: type of the image
  • image_subcat: image subtype, if present (outdoor, for example)
  • d1-d5: five descriptions in a sequence, corresponding to the image

splits.json is required if you want to use our original splits and, for example, exclude images labeled as extra, for which we have additionally collected description sequences, but did not include them in our analysis, described in the paper. However, these extra sequences are available in sequences.csv.

download_images.sh is the bash script, which will download ADE20k corpus into your repository.

Citation

If you find our data useful, please cite

Nikolai Ilinykh, Sina Zarrieß, and David Schlangen. 2019. Tell me more: A dataset of visual scene description sequences. In Proceedings of the 12th International Conference on Natural Language Generation, pages 152–157, Tokyo, Japan. Association for Computational Linguistics

This paper can be also found here.

If you want to use a short name for this corpus in your paper, please use IDS-ADE (for "image description sequences, for the ADE20k corpus").

About

A corpus of 5 turn image descriptions

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages