---
license: apache-2.0
dataset_info:
  features:
  - name: scene_id
    dtype: string
  - name: id
    dtype: string
  - name: target
    dtype: image
  - name: source
    dtype: string
  - name: source_image
    dtype: image
  - name: labels
    dtype:
      array2_d:
        shape:
        - 584
        - 876
        dtype: int32
  - name: cond_depths
    dtype:
      array2_d:
        shape:
        - 584
        - 876
        dtype: float32
  - name: extrin
    dtype:
      array2_d:
        shape:
        - 4
        - 4
        dtype: float32
  - name: intrin
    dtype:
      array2_d:
        shape:
        - 4
        - 4
        dtype: float32
  - name: depth
    dtype: string
  - name: item_id
    dtype: string
  - name: prompt
    dtype: string
  splits:
  - name: train
    num_bytes: 131813916499.633
    num_examples: 27799
  download_size: 73439626164
  dataset_size: 131813916499.633
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
task_categories:
- text-to-3d
---

This dataset is used in the paper [SceneCraft: Layout-Guided 3D Scene Generation](https://huggingface.co/papers/2410.09049).

# File information

The repository contains the following file information: