---
dataset_info:
  features:
  - name: scene
    dtype: string
  - name: image
    dtype: image
  - name: depth_map
    dtype: image
  - name: direction
    dtype: string
  - name: temprature
    dtype: int32
  - name: caption
    dtype: string
  splits:
  - name: train
    num_bytes: 20575644792.0
    num_examples: 12000
  download_size: 20108431280
  dataset_size: 20575644792.0
---
# VIDIT Dataset
This is a version of the [VIDIT dataset](https://github.com/majedelhelou/VIDIT) equipped for training ControlNet using depth maps conditioning.
VIDIT includes 390 different Unreal Engine scenes, each captured with 40 illumination settings, resulting in 15,600 images. The illumination settings are all the combinations of 5 color temperatures (2500K, 3500K, 4500K, 5500K and 6500K) and 8 light directions (N, NE, E, SE, S, SW, W, NW). Original image resolution is 1024x1024.
We include in this version only the training split containing only 300 scenes.
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 with varying direction
![varying direction](B_directions.gif)
## Examples with varying color temperature
![varying color temperature](B_illuminants.gif)

## Disclaimer
I do not own any of this data.