sdeval

Evaluation for stable diffusion model training

Installation

You can simply install it with pip command line from the official PyPI site.

pip install sdeval

If your operating environment includes a available GPU, you can use the following installation command to achieve higher performance:

pip install sdeval[gpu]

For more information about installation, you can refer to Installation.

Quick Start

CCIP Metrics

from sdeval.fidelity import CCIPMetrics

ccip = CCIPMetrics(images='/path/of/character/dataset')

# ccip score of one image
print(ccip.score('/path/of/one/image'))

# ccip score of a directory of images
print(ccip.score('/directory/of/images'))

Bikini Plus Metrics

from sdeval.controllability import BikiniPlusMetrics

# build bikini plus score metrics
bp = BikiniPlusMetrics(
    tag_blacklist=[
        'bangs', 'long_hair', 'blue_eyes', 'animal_ears', 'sleeveless',
        'breasts', 'grey_hair', 'medium_breasts'
    ]
)

# bp score of one image
# the image should contain a1111 webui's metadata of prompts
print(bp.score('/path/of/one/image'))

# bp score of a directory of images
# the images should contain a1111 webui's metadata of prompts
print(bp.score('/directory/of/images'))

AI Corrupt Metrics

from sdeval.corrupt import AICorruptMetrics

# build metrics
metrics = AICorruptMetrics()

# get ai corrupt score for one image file
print(metrics.score('/path/of/one/image'))

# get ai corrupt score of a directory of image files
print(metrics.score('/directory/of/images'))

# get ai corrupt score of list of images
print(metrics.score(['image1.png', 'image2.png']))