| --- |
| annotations_creators: |
| - crowdsourced |
| language: |
| - en |
| language_creators: |
| - found |
| license: |
| - cc-by-4.0 |
| multilinguality: |
| - monolingual |
| paperswithcode_id: seetrue |
| pretty_name: SeeTRUE |
| size_categories: |
| - 1K<n<10K |
| source_datasets: |
| - original |
| tags: |
| - image-captioning |
| - text-image-matching |
| task_ids: [] |
| extra_gated_prompt: "By clicking on “Access repository” below, you also agree that you are using it solely for research purposes, and that SeeTRUE should be used as a *TEST SET*, not as a training set, and especially not to train commercial chatbots. Do not hessitate to contact yonatanbitton@google.com if you have questions about this license." |
| --- |
| # Dataset Card for SeeTRUE |
|
|
| - [Dataset Description](#dataset-description) |
| - [Supported Tasks and Leaderboards](#supported-tasks-and-leaderboards) |
| - [Languages](#languages) |
| - [Dataset Structure](#dataset-structure) |
| - [Data Fields](#data-fields) |
| - [Data Splits](#data-splits) |
| - [Dataset Creation](#dataset-creation) |
| - [Licensing Information](#licensing-information) |
| - [Citation Information](#citation-information) |
|
|
| ## Dataset Description |
| The SeeTRUE dataset is a diverse benchmark for meta-evaluation of image-text alignment methods, covering the 4-way combinations of real and synthetic text-and-image pairs. It addresses limitations in current benchmarks, which mainly focus on natural images and often lack challenging negative captions. SeeTRUE allows to better assess the generalization abilities of text-image alignment models across various tasks. |
| We will add more datasets from SeeTRUE (e.g., COCO-Con and PickaPic-Con) upon data release. |
| Paper: https://arxiv.org/abs/2305.10400 |
| Website: https://wysiwyr-itm.github.io/ |
|
|
| ### Languages |
| The dataset supports English language. |
|
|
| ## Dataset Structure |
| ### Data Fields |
| - image: The name of the image file. |
| - text: The text description that matches with the image. |
| - label: The binary label. 1 if the text matches with the image, 0 otherwise. |
| - original_dataset_id: The ID of the dataset where the row originates from. |
| - dataset_source: The source of the dataset. |
| |
| ### Data Splits |
| SeeTRUE contains a single split: TEST, and should not be used for training. |
| |
| ## Dataset Creation |
| The dataset has been created by sourcing and matching images and text from multiple datasets. More information in the paper: https://arxiv.org/abs/2305.10400. |
| |
| ### Licensing Information |
| The dataset is under the CC-By 4.0 license. |
| |
| ### Citation Information |
| @article{yarom2023you, |
| title={What You See is What You Read? Improving Text-Image Alignment Evaluation}, |
| author={Yarom, Michal and Bitton, Yonatan and Changpinyo, Soravit and Aharoni, Roee and Herzig, Jonathan and Lang, Oran and Ofek, Eran and Szpektor, Idan}, |
| journal={arXiv preprint arXiv:2305.10400}, |
| year={2023} |
| } |
