Welcome to the Dataset Automaker Git repository housing an innovative and reliable Jupyter Notebook designed to facilitate the creation of datasets for training Stable Diffusion LoRAs.
Key Features:
-
Automated Anime Screencap Collection: Easily search and download anime screencaps from fancaps.net.
-
Advanced Data Refinement: Utilize the FiftyOne app and clip-vit-torch model to meticulously filter and curate your dataset for optimal quality.
-
Precise Face Detection: Leverage face detection models for accurate anime face identification (zymk9/yolov5_anime, ultralytics/yolov5).
-
Character Similarity Analysis: Calculate pairwise similarity distances between original and example faces, enabling targeted dataset curation with user-defined thresholds.
-
User-Guided Curation: Utilize the intuitive FiftyOne app for manual dataset refinement, ensuring precision.
-
Intelligent Tagging: Effortlessly tag results using code by kohya-colab and sd-scripts, enhancing organization and analysis.
-
Seamless Export: Conveniently zip and download your curated dataset.
Local:
- Clone or download the repository to your local machine.
- Launch Jupyter Notebook and open the Dataset_Automaker.ipynb notebook.
- Follow the comprehensive guide within the notebook to harness the power of automated data generation.
Google colab:
- Open Dataset_Automaker.ipynb in Google colab.
- Follow the guide in the notebook
Civitai: Maximax67
Telegram: @Maximax67
Github: Maximax67
Gmail: maximax6767@gmail.com