dataset_info:
features:
- name: filename
dtype: image
- name: short_description
dtype: string
- name: long_description
dtype: string
- name: link
dtype: string
configs:
- config_name: default
data_files:
- split: test
path:
- prompts_and_classes.json
- images.zip
DesignBench Dataset
Data Info
The dataset includes 30 images containing different objects and scenes.
The images have been sourced from www.unsplash.com, www.pexels.com, www.freepik.com (the licenses for the images are listed below).
Out of the 30 images: 8 contain animals, 9 images contain humans or partial human figures, 15 images contain only inanimate objects and 2 contain only a scene. The dataset contains a variable number of subjects (1-8) per image. Images are captured in different conditions, environments and under different angles.We include a file dataset/prompts_and_classes.json which contains two types of prompts per image: a lengthy detailed prompt and a short concise prompt lacking details. These are the prompts used in the paper for all experiments using DesignBench. The images have been sourced from www.unsplash.com, www.pexels.com, www.freepik.com.
Data Access
designbench/prompts_and_classes.json file contains a list of all the image names, reference links to the images, and a short and long prompt per image. The images have been cropped from their original form to directly download the cropped version of the photos that was used in the paper visit designbench/images/
Licenses
Unsplash: Unsplash grants you an irrevocable, nonexclusive, worldwide copyright license to download, copy, modify, distribute, perform, and use images from Unsplash for free, including for commercial purposes, without permission from or attributing the photographer or Unsplash. This license does not include the right to compile images from Unsplash to replicate a similar or competing service.
Pexels: All photos and videos uploaded on Pexels are licensed under the Pexels license. This means you can use them for free for personal and commercial purposes without attribution. For more information read the following questions in this guide, or refer to our license page or our Terms of Service.
FreePik: Freepik and Flaticon PDF licenses allow you to use our resources without crediting the author and these will remain active even when your Premium or Premium+ subscription has expired.Please remember to download the license PDF files and keep them in a safe place. We recommend you to download them right after downloading the file.https://support.freepik.com/s/article/How-to-download-Freepik-premium-licenses?language=en_US
Citing this work
If you find this code useful, please cite our paper
@misc{hahn2024proactive,
title={Proactive Agents for Multi-Turn Text-to-Image Generation Under Uncertainty},
author={Meera Hahn and Wenjun Zeng and Nithish Kannen and Rich Galt and Kartikeya Badola and Been Kim and Zi Wang},
year={2024},
eprint={2412.06771},
archivePrefix={arXiv},
primaryClass={cs.AI}
}
License
Copyright 2024 DeepMind Technologies Limited
All software is licensed under the Apache License, Version 2.0 (Apache 2.0); you may not use this file except in compliance with the Apache 2.0 license. You may obtain a copy of the Apache 2.0 license at: https://www.apache.org/licenses/LICENSE-2.0
All other materials are licensed under the Creative Commons Attribution 4.0 International License (CC-BY). You may obtain a copy of the CC-BY license at: https://creativecommons.org/licenses/by/4.0/legalcode
Unless required by applicable law or agreed to in writing, all software and materials distributed here under the Apache 2.0 or CC-BY licenses are distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the licenses for the specific language governing permissions and limitations under those licenses.
This is not an official Google product.
Contact
Please direct any questions to wangzi@google.com