scipostlayouttree / README.md
nielsr's picture
nielsr HF Staff
Improve dataset card: Add paper link, task categories, correct size, and sample usage
550a515 verified
|
raw
history blame
1.97 kB
metadata
language:
  - en
license: cc
size_categories:
  - 1K<n<10K
task_categories:
  - OTHER
tags:
  - scientific-posters
  - layout-analysis
  - document-structure-analysis

SciPostLayoutTree: A Dataset for Structural Analysis of Scientific Posters

This repository contains the SciPostLayoutTree dataset, introduced in the paper "SciPostLayoutTree: A Dataset for Structural Analysis of Scientific Posters".

SciPostLayoutTree is a dataset of approximately 8,000 scientific posters annotated with reading order and parent-child relations. It is designed to facilitate research into structural analysis of scientific posters, which play a vital role in academic communication. The dataset specifically addresses challenges related to spatially complex relations, including upward, horizontal, and long-distance relationships, making it a valuable resource for building structure-aware interfaces.

For detailed dataset construction, experimental setup, and reproduction steps, please refer to the comprehensive instructions provided in the GitHub repository.

Sample Usage

You can visualize tree annotations using the script provided in the associated GitHub repository. First, install the necessary Python packages:

pip install opencv-python numpy matplotlib tqdm

Then, execute the visualization script:

python visualize_annotation.py

Please ensure you have completed the Dataset Construction steps outlined in the GitHub repository to prepare the data and annotation files before running the visualization.