The viewer is disabled because this dataset repo requires arbitrary Python code execution. Please consider
removing the loading script and relying on automated data support (you can use convert_to_parquet from the datasets library). If this is not possible, please open a discussion for direct help.
All pictures from pixel with tags "Keqing", "keqing", "Keqing", "KeQing" and "Keqing (Yuanshin)"
AI Generator has been marked, classified by number of collections, R18 content removed (please eat with confidence, Mommy Qingqing)
The work ID, author ID, and author name have been tagged. If you like it, you can go to the author's collection to reward and support.
The original dataset size is 81.6GB, with 33,480 images. Currently, the deduplication experiment for the Keqing dataset based on deep learning has been completed. In the future, there may be additions such as emotion classification, clothing classification, image super-resolution, image denoising, generating videos from images using Stable Video Diffusion, generating images from labels using Stable Diffusion, and cloning Keqing's voice based on GPT-SoVITS. A scoring system for Keqing images will be developed, weighted by the number of collections.
After deduplication, the dataset consists of 28,400 images. After being compressed into volumes using 7z, the size is 67.9GB, and the password is Keqing's birthday. Please use 7z to open it.
- Downloads last month
- 6