omron-sinicx/scipostlayouttree

---
license: cc
language:
- en
size_categories:
- 1K<n<10K
task_categories:
- other
tags:
- scientific-posters,
- layout-analysis,
- document-structure-analysis
---

This is the dataset repository of a paper ''SciPostLayoutTree: A Dataset for Structural Analysis of Scientific Posters'' (https://arxiv.org/abs/2511.18329).
SciPostLayoutTree is a dataset of about 8,000 scientific posters annotated with reading order information and parent child relations, represented as DFS ordered trees. The dataset extends SciPostLayout by adding region level structural annotations that cover eight layout categories including Title, Author Info, Section, Text, List, Figure, Table, and Caption. Scientific posters exhibit dense and diverse spatial arrangements, and SciPostLayoutTree reflects this property through frequent upward, horizontal, and long distance relations that are not commonly observed in document based datasets. These characteristics make the dataset useful for training and evaluating models designed to understand the hierarchical organization of visually rich poster layouts.
For dataset usage and reproduction steps, please refer to the instructions provided in the following repository: https://github.com/omron-sinicx/scipostlayouttree