README.md

# Anime Head Detection Models

This repository contains three object detection models based on YOLOv5l and YOLOv8l architectures. The images in the dataset were taken from anime and art, with manga pages added to AnimeHeadsv3.

## Versions

### AnimeHeadsv3

- Description: The third and last version of the project based on YOLOv8l. The training set contains 8037 images.
- Augmentation parameters: 
    - Outputs per training example: 3
    - Flip: Horizontal
    - Saturation: Between -40% and +40%
    - Blur: Up to 4px
    - Noise: Up to 4% of pixels
- Example images:

    ![Example image 1](https://huggingface.co/nyuuzyou/AnimeHeads/resolve/main/examples/1v3.jpg)
    ![Example image 2](https://huggingface.co/nyuuzyou/AnimeHeads/resolve/main/examples/2v3.jpg)
    ![Example image 3](https://huggingface.co/nyuuzyou/AnimeHeads/resolve/main/examples/3v3.jpg)
    
- Model file: `animeheadsv3.weight`

### AnimeHeadsv2

- Description: The second version of the project based on YOLOv8l. The training set contains 4632 images.
- Augmentation parameters: 
    - Outputs per training example: 3
    - Blur: Up to 2.75px
    - Noise: Up to 5% of pixels
- Example images:

    ![Example image 1](https://huggingface.co/nyuuzyou/AnimeHeads/resolve/main/examples/1v2.jpg)
    ![Example image 2](https://huggingface.co/nyuuzyou/AnimeHeads/resolve/main/examples/2v2.jpg)
    ![Example image 3](https://huggingface.co/nyuuzyou/AnimeHeads/resolve/main/examples/3v2.jpg)
- Model file: `animeheadsv2.weight`

### AnimeHeads

- Description: The first version of the project based on YOLOv5l. The training set contains 886 images.
- Augmentation parameters: None
- Example images:

    ![Example image 1](https://huggingface.co/nyuuzyou/AnimeHeads/resolve/main/examples/1v1.jpg)
    ![Example image 2](https://huggingface.co/nyuuzyou/AnimeHeads/resolve/main/examples/2v1.jpg)
    ![Example image 3](https://huggingface.co/nyuuzyou/AnimeHeads/resolve/main/examples/3v1.jpg)
- Model file: `animeheads.weight`