DALLE-URBAN: Capturing the Urban Design Expertise of Large Text to Image Transformers

This paper investigates the capabilities and biases of text-to-image methods as they apply to the built environment in detail. We evaluate how large generative models such as DALL-E capture urban design knowledge, examining both the potential and limitations of these models for generating imagery of urban environments.

Download paper here

Recommended citation:

S. Seneviratne, D. Senanayake, S. Rasnayaka, R. Vidanaarachchi and J. Thompson, “DALLE-URBAN: Capturing the Urban Design Expertise of Large Text to Image Transformers,” 2022 International Conference on Digital Image Computing: Techniques and Applications (DICTA), 2022, pp. 1-9, doi: 10.1109/DICTA56598.2022.10034603.

BibTeX

@inproceedings{seneviratne2022dalle,
  title={DALLE-URBAN: Capturing the urban design expertise of large text to image transformers},
  author={Seneviratne, Sachith and Senanayake, Damith and Rasnayaka, Sanka and Vidanaarachchi, Rajith and Thompson, Jason},
  booktitle={2022 International Conference on Digital Image Computing: Techniques and Applications (DICTA)},
  pages={1--9},
  year={2022},
  organization={IEEE}
}