Unverified Commit 1a987934 authored by narugo1992's avatar narugo1992 Committed by GitHub
Browse files

Merge pull request #27 from deepghs/dev/cls 

dev(narugo): use dist model for anime classification
parents ab56db74 00b3b087

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

+599 −395

File changed.

Preview size limit exceeded, changes collapsed.

imgutils/validate/classify.py

+5 −3
Original line number Diff line number Diff line
@@ -35,10 +35,12 @@ _MODEL_NAMES = [ 
    'caformer_s36',

    'caformer_s36_plus',

    'mobilenetv3',

    'mobilenetv3_dist',

    'mobilenetv3_sce',

    'mobilenetv3_sce_dist',

    'mobilevitv2_150',

]

_DEFAULT_MODEL_NAME = 'mobilenetv3_sce'

_DEFAULT_MODEL_NAME = 'mobilenetv3_sce_dist'





@lru_cache()
@@ -77,7 +79,7 @@ def anime_classify_score(image: ImageTyping, model_name: str = _DEFAULT_MODEL_NA 
        Predict the class of the given image, return the score with as a dict object.



    :param image: Image to classify.

    :param model_name: Model to use. Default is ``mobilenetv3_sce``. All available models are listed

    :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.
@@ -120,7 +122,7 @@ def anime_classify(image: ImageTyping, model_name: str = _DEFAULT_MODEL_NAME) -> 
        Predict the class of the given image, return the class and its score.



    :param image: Image to classify.

    :param model_name: Model to use. Default is ``mobilenetv3_sce``. All available models are listed

    :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.

test/testfile/anime_cls/illustration/4726239.jpg

deleted100644 → 0
−45.6 KiB
Loading image diff...