---
dataset_info:
  features:
  - name: indices
    dtype: string
  - name: plans
    dtype: image
  - name: walls
    dtype: image
  - name: colors
    dtype: image
  - name: footprints
    dtype: image
  - name: plan_captions
    dtype: string
  splits:
  - name: train
    num_bytes: 3999080609
    num_examples: 12000
  download_size: 2497201625
  dataset_size: 3999080609
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
tags:
- architecture
size_categories:
- 10K<n<100K
---
# Dataset Card for "pseudo-floor-plan-12k"
This is a experimental dataset for training conditional text-to-image models for generating compiled architectural floor plans (rather than simple walls/rooms segmentaion) with certain degree of controls(walls, rooms, footprint boundrays, etc.).

The source images are procedurally generated through Grasshopper with PlanFinder plugin, the quality of the generated plans are not as reliable as the real architectural plans, please use with **caution**.

The collection and processing of real floor plans data are difficcult and time consuming due to the limited resource under this topic, especially for the compiled plans with sufficient detials. 

The "real floor plans" dataset is under development.

[More Information needed](https://github.com/huggingface/datasets/blob/main/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)