BangumiBase

这是怎麽？
将动漫视频下载到我们的 GPU 集群后，我们采用各种计算机视觉算法来提取帧、裁剪并根据角色特征对其进行聚类 。然后将这些处理后的帧上传到此处，以减少角色图像所需的手动排序工作。

该存储库中的数据将经过自动二次处理以消除噪音，然后将其打包并上传到 CyberHarem。然后，它将集成到用于训练角色 LoRA 的自动化管道中。

当前动漫数据库（不断更新）
最后更新于：2025-09-26 09：07，共收录动漫808部，图片4295852张，共47537个集群。

license: Apache License 2.0
image:
image-classification:
size_scale:
- 10k-1m
tags: []

#### 下载方法
:modelscope-code[]{type="sdk"}
:modelscope-code[]{type="git"}

What's this?
After downloading anime videos to our GPU cluster, we employ various computer vision algorithms to extract frames, crop them, and cluster them based on character features. These processed frames are then uploaded here to reduce the manual sorting work required for character images.

The data in this repository will undergo automatic secondary processing to remove noise, then be packaged and uploaded to CyberHarem. It will then be integrated into the automated pipeline for training character LoRAs.

### Current Anime Database (continuously updated)
Last updated: 2025-09-26 09:07, containing a total of 808 anime series, 4,295,852 images, and 47,537 clusters.

License: Apache License 2.0
image:
image-classification:
size_scale:
- 10k-1m
tags: []

#### Download Methods
:modelscope-code[type="sdk"]
:modelscope-code[type="git"]