Hugging Face's logo

Datasets:
MujicaLab
/
MyGO-Mujica-VideoFrames

Modalities:
Video
Languages:
English
Size:
1K - 10K
Libraries:
Datasets
License:
Dataset card Data Studio Files Community
1

You need to agree to share your contact information to access this dataset

This repository is publicly accessible, but you have to accept the conditions to access its files and content.

Log in or Sign Up to review the conditions and access this dataset content.

BanG Dream! MyGO & Ave Mujica - Video Frames Dataset / BanG Dream! MyGO 与 Ave Mujica - 视频帧数据集

This dataset contains raw videos, scene-split video clips, and extracted frames from the anime series "BanG Dream! It's MyGO!!!!!" and "BanG Dream! Ave Mujica".

本数据集包含动画系列《BanG Dream! It's MyGO!!!!!》和《BanG Dream! Ave Mujica》的原始视频、按场景分割的视频片段以及提取的视频帧。

Content / 数据集内容

  1. Raw Videos: Original video files for both series, along with scene timestamps. / 原始视频: 两部番剧的原始视频文件及对应的场景时间戳。
  2. Split Videos: Videos split into individual scenes based on scene detection. / 分割视频: 使用场景检测将视频分割成的独立场景片段。
  3. Extracted Frames: Deduplicated individual frames extracted from the videos, saved as WebP images, along with corresponding feature vectors. / 提取帧: 从视频中提取并去重的单帧图像（WebP 格式）及其对应的特征向量。

Directory Structure / 目录结构

The dataset is organized into two main folders, one for each series:

数据集被组织成两个主文件夹，分别对应两个系列：

  • mygo/ : Contains data for "BanG Dream! It's MyGO!!!!!" / 包含《BanG Dream! It's MyGO!!!!!》的数据。
  • ave_mujica/ : Contains data for "BanG Dream! Ave Mujica" / 包含《BanG Dream! Ave Mujica》的数据。

Within each of these series folders, you will find the following subdirectories:

在每个系列文件夹内，您会找到以下子目录：

  1. raw_videos/

    • Contains the original video files, typically in .mkv format (e.g., 01.mkv, 02.mkv, ...). / 包含原始视频文件，通常为 .mkv 格式（例如 01.mkv, 02.mkv, ...）。
    • Includes scene list files generated by PySceneDetect, named like XX-Scenes.csv (e.g., 01-Scenes.csv), detailing the start and end times/frames for each detected scene in the corresponding video file. / 包含由 PySceneDetect 生成的场景列表文件，命名格式如 XX-Scenes.csv（例如 01-Scenes.csv），其中详细记录了对应视频文件中每个检测到的场景的开始和结束时间/帧。

  2. split_videos/

    • Contains subfolders for each episode, named 01, 02, ..., 13. / 包含每个剧集的子文件夹，命名为 01, 02, ..., 13。
    • Inside each episode folder (e.g., 01/), you'll find the video clips corresponding to the detected scenes, named sequentially like scene_0001.mp4, scene_0002.mp4, ..., scene_NNNN.mp4. These are the results of splitting the raw videos based on the scene lists. / 在每个剧集文件夹（例如 01/）内，您会找到与检测到的场景相对应的视频片段，按顺序命名，如 scene_0001.mp4, scene_0002.mp4, ..., scene_NNNN.mp4。这些是根据场景列表分割原始视频的结果。

  3. extracted_frames/

    • Contains compressed archive files for each episode, named 01.tar, 02.tar, ..., 13.tar. / 包含每个剧集的压缩归档文件，命名为 01.tar, 02.tar, ..., 13.tar。
    • Each .tar file contains the extracted and deduplicated frames for that episode. / 每个 .tar 文件包含对应剧集的已提取并去重的帧。
      • Frames are saved as WebP images (.webp) with 90% quality. Deduplication was performed using methods including LPIPS to remove visually similar consecutive frames. / 帧以 90% 质量的 WebP 图像 (.webp) 格式保存。使用了包括 LPIPS 在内的多种方法进行去重，以移除视觉上相似的连续帧。
      • Image filenames follow the pattern EP_NNNN.webp, where EP is the zero-padded episode number (e.g., 01) and NNNN is the sequential frame number within that episode (e.g., 01_0001.webp). / 图像文件名遵循 EP_NNNN.webp 格式，其中 EP 是补零的剧集编号（例如 01），NNNN 是该剧集内帧的序号（例如 01_0001.webp）。
      • For each .webp image, there is a corresponding .npy file with the same base name (e.g., 01_0001.npy). / 对于每个 .webp 图像，都有一个同名的 .npy 文件（例如 01_0001.npy）。
      • These .npy files contain feature vectors extracted from the corresponding image using the SmilingWolf/wd-swinv2-tagger-v3 model. / 这些 .npy 文件包含使用 SmilingWolf/wd-swinv2-tagger-v3 模型从对应图像中提取的特征向量。

Data Sources / 数据来源

Processing Details / 处理细节

  • Video Quality: Both the raw videos and the split video clips are kept in their original quality without re-compression. / 视频质量: 原始视频和分割后的视频片段均未经过二次压缩，保留了原始素材质量。
  • Frame Format: Extracted frames are saved in WebP format with 90% quality. / 帧格式: 提取的帧以 90% 质量的 WebP 格式保存。
  • Tools Used / 使用工具:

License / 许可证

Important Notice: The original video content ("BanG Dream! It's MyGO!!!!!" and "BanG Dream! Ave Mujica") is copyrighted by its respective owners . This dataset, containing derived materials (clips and frames), is provided strictly for non-commercial research, personal study, and educational purposes only, potentially falling under fair use or similar exceptions depending on your jurisdiction.

重要提示： 原始视频内容（《BanG Dream! It's MyGO!!!!!》和《BanG Dream! Ave Mujica》）的版权归其各自所有者所有。本数据集包含衍生素材（视频片段和帧），严格仅用于非商业研究、个人学习和教育目的，根据您所在司法管辖区的不同，可能属于合理使用或类似例外情况。

No Rights Granted: The distribution of this dataset does not grant any license or rights to the underlying copyrighted works. Users are solely responsible for ensuring their use of this dataset complies with applicable copyright laws and respects the rights of the original copyright holders.

未授予任何权利： 本数据集的分发 不 授予对基础受版权保护作品的任何许可或权利。用户需自行负责确保其对本数据集的使用符合适用的版权法律，并尊重原始版权所有者的权利。

Use At Your Own Risk: By using this dataset, you acknowledge and agree to these terms. The creators of this dataset assume no liability for misuse. Please support the official releases.

使用风险自负： 使用本数据集即表示您承认并同意这些条款。本数据集的创建者对滥用不承担任何责任。请支持官方正版发布。

Downloads last month
78