metadata

language:
  - en
license: cc-by-4.0
size_categories:
  - 10K<n<100K
task_categories:
  - image-to-text
pretty_name: RICO Screen Annotations
tags:
  - screens
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
      - split: valid
        path: data/valid-*
      - split: test
        path: data/test-*
dataset_info:
  features:
    - name: screen_id
      dtype: string
    - name: screen_annotation
      dtype: string
    - name: file_name
      dtype: string
    - name: image
      dtype: image
  splits:
    - name: train
      num_bytes: 1684182938.288
      num_examples: 15548
    - name: valid
      num_bytes: 240141824.938
      num_examples: 2311
    - name: test
      num_bytes: 452100376.53
      num_examples: 4217
  download_size: 1880458708
  dataset_size: 2376425139.756

Dataset Card for RICO Screen Annotations

This is a standardization of Google's Screen Annotation dataset on a subset of RICO screens, as described in their ScreenAI paper. Unlike the original, this version transforms integer-based bounding boxes into floating-point-based bounding boxes of 2 decimal precision.

Dataset Details

Dataset Description

This is an image-to-text annotation format first proscribed in Google's ScreenAI paper. The idea is to standardize an expected text output that is reasonable for the model to follow, and fuses together things like element detection, referring expression generation/recognition, and element classification.

Curated by: Google Research
Language(s) (NLP): English
License: CC-BY-4.0

Dataset Sources

Repository: google-research/screen_annotation
Paper [optional]: ScreenAI

Uses

Direct Use

Pre-training of multimodal models to better understand screens.

Dataset Structure

screen_id: Screen ID in the RICO dataset
screen_annotation: Target output string
image: The RICO screenshot

Dataset Creation

Curation Rationale

The Screen Annotation dataset consists of pairs of mobile screenshots and their annotations. The mobile screenshots are directly taken from the publicly available Rico dataset. The annotations are in text format, and contain information on the UI elements present on the screen: their type, their location, the text they contain or a short description. This dataset has been introduced in the paper ScreenAI: A Vision-Language Model for UI and Infographics Understanding and can be used to improve the screen understanding capabilities of multimodal (image+text) models.

Citation

BibTeX:

@misc{baechler2024screenai,
      title={ScreenAI: A Vision-Language Model for UI and Infographics Understanding},
      author={Gilles Baechler and Srinivas Sunkara and Maria Wang and Fedir Zubach and Hassan Mansoor and Vincent Etter and Victor Cărbune and Jason Lin and Jindong Chen and Abhanshu Sharma},
      year={2024},
      eprint={2402.04615},
      archivePrefix={arXiv},
      primaryClass={cs.CV}
}

Dataset Card Authors

Hunter Heidenreich, Roots Automation

Dataset Card Contact

hunter "dot" heidenreich AT rootsautomation DOT com