anime-seg 0.0.3

Anime Character Segmentation with DINOv2

AnimeSeg

Anime Character Segmentation using DINOv2 + U-Net++ with LoRA fine-tuning.

Installation

pip install -e .

Usage

from anime_seg import AnimeSegPipeline

# Initialize pipeline (auto-downloads latest model from Hugging Face)
pipe = AnimeSegPipeline()

# Run segmentation
mask = pipe("path/to/image.jpg")

# Save result
mask.save("output.png")

Advanced Usage

# Specify custom repo or filename
pipe = AnimeSegPipeline(
    repo_id="suzukimain/AnimeSeg",
    filename="models/anime_seg_dinov2_large_v1.safetensors",
    device="cuda"  # or "cpu"
)

# Use PIL Image
from PIL import Image
img = Image.open("image.jpg")
mask = pipe(img)

Model Files

Models should follow the naming convention:

models/anime_seg_{architecture}_{size}_v{version}.safetensors

Example:

• models/anime_seg_dinov2_large_v1.safetensors
• models/anime_seg_dinov2_base_v2.safetensors

Segmentation Classes

• Background
• Skin
• Face
• Hair (main)
• Hair (thin)
• Eyes (left/right)
• Eyebrows (left/right)
• Nose
• Mouth
• Clothes
• Unknown