Commit 7813255c authored by narugo1992's avatar narugo1992
Browse files

dev(narugo): regenerate plots, test skip

parent 43419774

docs/source/api_doc/validate/index.rst

+1 −0
Original line number Diff line number Diff line
@@ -13,4 +13,5 @@ imgutils.validate 
    classify

    color

    monochrome

    rating

    truncate

docs/source/api_doc/validate/rating.plot.py

+1 −1
Original line number Diff line number Diff line
@@ -10,5 +10,5 @@ if __name__ == '__main__': 
        *natsorted(glob.glob(os.path.join('rating', 'safe', '*.jpg'))),

        *natsorted(glob.glob(os.path.join('rating', 'r15', '*.jpg'))),

        *natsorted(glob.glob(os.path.join('rating', 'r18', '*.jpg'))),

        columns=4, figsize=(10.5, 10),

        columns=4, figsize=(10, 12),

    )

docs/source/api_doc/validate/rating.plot.py.svg

deleted100644 → 0
+0 −1087

File deleted.

Preview size limit exceeded, changes collapsed.

docs/source/api_doc/validate/rating.rst

0 → 100644
+21 −0
Original line number Diff line number Diff line 
imgutils.validate.rating

=============================================



.. currentmodule:: imgutils.validate.rating



.. automodule:: imgutils.validate.rating





anime_rating_score

-----------------------------



.. autofunction:: anime_rating_score







anime_rating

-----------------------------



.. autofunction:: anime_rating

imgutils/validate/rating.py

+14 −4
Original line number Diff line number Diff line
@@ -14,6 +14,16 @@ Overview: 


    The models are hosted on

    `huggingface - deepghs/anime_rating <https://huggingface.co/deepghs/anime_rating>`_.



    .. note::

        Please note that the classification of ``safe``, ``r15``, and ``r18`` types does not have clear boundaries,

        making it challenging to clean the training data. As a result, there is no strict ground truth

        for the rating classification problem. The judgment functionality provided by the current module

        is intended as a quick and rough estimation.



        **If you require an accurate filtering or judgment function specifically for R-18 images,

        it is recommended to consider using object detection-based methods**,

        such as using :func:`imgutils.detect.censor.detect_censors` to detect sensitive regions as the basis for judgment.

"""

import json

from functools import lru_cache
@@ -82,12 +92,12 @@ def _raw_anime_rating(image: ImageTyping, model_name: str = _DEFAULT_MODEL_NAME) 
def anime_rating_score(image: ImageTyping, model_name: str = _DEFAULT_MODEL_NAME) -> Dict[str, float]:

    """

    Overview:

        Predict the class of the given image, return the score with as a dict object.

        Predict the rating of the given image, return the score with as a dict object.



    :param image: Image to rating.

    :param model_name: Model to use. Default is ``mobilenetv3_sce_dist``. All available models are listed

        on the benchmark plot above. If you need better accuracy, just set this to ``caformer_s36_plus``.

    :return: A dict with classes and scores.

    :return: A dict with ratings and scores.

    """

    output = _raw_anime_rating(image, model_name)

    values = dict(zip(_open_anime_rating_labels(model_name), map(lambda x: x.item(), output[0])))
@@ -97,12 +107,12 @@ def anime_rating_score(image: ImageTyping, model_name: str = _DEFAULT_MODEL_NAME 
def anime_rating(image: ImageTyping, model_name: str = _DEFAULT_MODEL_NAME) -> Tuple[str, float]:

    """

    Overview:

        Predict the class of the given image, return the class and its score.

        Predict the rating of the given image, return the class and its score.



    :param image: Image to rating.

    :param model_name: Model to use. Default is ``mobilenetv3_sce_dist``. All available models are listed

        on the benchmark plot above. If you need better accuracy, just set this to ``caformer_s36_plus``.

    :return: A tuple contains the class and its score.

    :return: A tuple contains the rating and its score.



    """

    output = _raw_anime_rating(image, model_name)[0]