--- license: mit --- ## 📚 Dataset Summary This dataset features **1,941 anime character images**, neatly organized into **322 folders**, each representing a different anime series 🎌. - 📦 **Size of downloaded files**: 152 MB - 🪄 **Size of auto-converted Parquet files**: 151 MB - 📊 **Split**: Train only - 🎭 **Classes**: 322 unique anime titles Perfect for **image classification**, **anime recommendation systems**, and **visual style analysis**! 🎨✨ --- ## 🏆 Supported Tasks - 🖼️ **Image Classification**: Predict the anime title based on a character image. - 🔍 **Visual Similarity Retrieval**: Find visually similar anime styles. - 🛠️ **Model Fine-tuning**: Customize vision models (ViT, ConvNeXt, etc.) for anime domains. --- ## 🗂️ Dataset Structure ### 📁 Directory Layout ``` Anime_Characters/ ├── dataset/ │ ├── .hack__G.U. Returner/ │ │ ├── Haseo.png │ │ ├── Ovan.png ... │ ├── Air Gear/ │ │ ├── Onigiri.png ... │ ├── Bleach/ │ │ ├── Ichigo Kurosaki.png ... │ └── (other anime folders...) ├── anime_characters_dataset.json ├── .gitignore └── README.md ``` Each folder inside `dataset/` represents an **anime class**, and contains images of characters from that anime. 📸 --- ## 📝 Metadata (anime_characters_dataset.json) The file `anime_characters_dataset.json` contains a list of all images and their associated labels. --- ## 🔢 Data Fields - `image_path` ➡️ Path to the character image. - `label` ➡️ Anime series title (class label). --- ## 📊 Data Splits | Split | Number of Images | |------|------------------| | Train | 1,941 | 🔔 *Note: No official validation/test split is provided — users can create their own!* --- ## 🚀 Usage Example ```python import json # Load the metadata with open('anime_characters_dataset.json', 'r') as f: data = json.load(f) # Example usage print(data[0]) ``` --- ## ⚖️ Licensing ⚠️ **Important**: This dataset may include content protected by copyright. It is intended strictly for **educational, research, or non-commercial use** under fair use laws. Always verify before commercial usage! 📜 --- ## 📌 Additional Notes - The dataset includes **raw images** and **JSON metadata**. - Images have been auto-converted to **Parquet** format for Hugging Face Dataset Viewer. - 📂 Anime_Characters: - `dataset/` - `anime_characters_dataset.json` - `.gitignore` - `README.md` ---