---
dataset_info:
  features:
  - name: image
    dtype: image
  - name: depth_map
    dtype: image
  - name: scene
    dtype: string
  - name: caption
    dtype: string
  - name: state
    dtype: string
  splits:
  - name: train
    num_bytes: 11835627985.25
    num_examples: 5550
  download_size: 12139477164
  dataset_size: 11835627985.25
---
# A Dataset of Flash and Ambient Illumination Pairs from the Crowd 

This is a version of the [A Dataset of Flash and Ambient Illumination Pairs from the Crowd](http://yaksoy.github.io/flashambient/) dataset equipped for training ControlNet using depth maps conditioning.
The dataset includes 2775 pairs of flash light and ambient light images. It includes images of people, shelves, plants, toys, rooms and objects.

Captions were generated using the [BLIP-2, Flan T5-xxl](https://huggingface.co/Salesforce/blip2-flan-t5-xxl) model.
Depth maps were generated using the [GLPN fine-tuned on NYUv2 ](https://huggingface.co/vinvino02/glpn-nyu) model.

## Examples
![Examples](faiTeaser.jpg)
## Disclaimer
I do not own any of this data.