Genius-Society/vtuber_emojis

Intro

This dataset is a multi-class VTuber chat sticker dataset constructed for classification tasks, focusing on image content centered around anime-style virtual avatars. It includes chat stickers widely used in livestream interactions, fan communities, and bullet-screen culture. The dataset is meticulously annotated at the individual VTuber level, enabling research in image classification, cross-modal retrieval, identity recognition, and visual feature learning. All data have been systematically curated and deduplicated, preserving key visual characteristics of chat stickers such as character illustrations, exaggerated facial expressions, overlaid text, and stylized design elements, while ensuring clear category boundaries and semantic consistency in annotation. The dataset can be used to train and evaluate computer vision models for virtual streamer identity recognition, and also provides foundational data support for research in anime-style visual culture, digital character dissemination, and multimodal affective understanding.

Usage

from datasets import load_dataset

ds = load_dataset(
    "Genius-Society/vtuber_emojis",
    name="default",
    split="train",
    cache_dir="./__pycache__",
)
for i in ds:
    print(i)

Maintenance

GIT_LFS_SKIP_SMUDGE=1 git clone git@hf.co:datasets/Genius-Society/vtuber_emojis
cd vtuber_emojis

Mirror

https://www.modelscope.cn/datasets/Genius-Society/vtuber_emojis