Skip to content

Latest commit

 

History

History
30 lines (20 loc) · 932 Bytes

README.md

File metadata and controls

30 lines (20 loc) · 932 Bytes

StyleTokenizer

This is the repo for ECCV 2024 accepted paper StyleTokenizer: Defining Image Style by a Single Instance for Controlling Diffusion Models.

Style30K dataset

The dataset can be found Here (pwd:ydve). The style category of each image can be obtained in the following way,

style_code = image_file_name.split('____')[0]

Additionally, if you are interested in more information about the style, you can find the style name according to the given 'style_mapper.json'

Comparison with other methods

Acknowledgement

Some images in dataset Style30K are from LAION-5B.