license: apache-2.0
AnyControl: Create Your Artwork with Versatile Control on Text-to-Image Generation
Yanan Sun, Yanchen Liu, Yinhao Tang, Wenjie Pei and Kai Chen*
Shanghai AI Laboratory
Dataset Card
We provide a benchmark for evaluating the model performance of unaligned multi-control image synthesis. The structure of COCO-UM is as follows,
COCO-UM
├── val2017
├── inpaint
├── mask
For each image in val2017, we pick an object in the image as the foreground object with mask and inpaint the holed background image with PowerPaint, as shown in the following figure.
The corresponding mask and inpainted results are stored in mask and inpaint folders, respectively.
When conducting evaluation on multi-control image synthesis, you can provide one or more control signals (e.g., pose or canny) for the foreground object (the masked region),
while utilize the other control signals (e.g. depth or segmentation) for the background region (the unmasked region).
License and Citation
All models and assets are under the Apache 2.0 license unless specified otherwise.
If this work is helpful for your research, please consider citing the following BibTeX entry.
@misc{sun2024anycontrol,
title={AnyControl: Create your artwork with versatile control on text-to-image generation},
author={Sun, Yanan and Liu, Yanchen and Tang, Yinhao and Pei, Wenjie and Chen, Kai},
booktitle={ECCV},
year={2024}
}