---
dataset_info:
  features:
  - name: image
    dtype: image
  - name: conditioning_image
    dtype: image
  - name: text
    dtype: string
  splits:
  - name: train
    num_bytes: 60051214963.0
    num_examples: 118285
  download_size: 60053644708
  dataset_size: 60051214963.0
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
---


I cropped the Coco 2017 dataset to 512 and created a depth image with DepthAnything v2.
https://cocodataset.org/#download