This is the image base of bangumi Blend S, we detected 16 characters, 1863 images in total. The full dataset is here.
Please note that these image bases are not guaranteed to be 100% cleaned, they may be noisy actual. If you intend to manually train models using this dataset, we recommend performing necessary preprocessing on the downloaded dataset to eliminate potential noisy samples (approximately 1% probability).
Here is the characters' preview:… See the full description on the dataset page:
https://huggingface.co/datasets/BangumiBase/blends.
