dev(narugo): save it (988b1d66) · Commits · git-mirror / Imgutils

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

+5 −14

Original line number	Diff line number	Diff line
		import os
		import random

		from huggingface_hub import HfFileSystem

		from benchmark import BaseBenchmark, create_plot_cli
		from imgutils.generic.classify import _open_models_for_repo_id
		from imgutils.validate import anime_portrait
		from imgutils.validate.portrait import _REPO_ID

		hf_fs = HfFileSystem()

		_REPOSITORY = 'deepghs/anime_portrait'
		_MODEL_NAMES = [
		os.path.relpath(file, _REPOSITORY).split('/')[0] for file in
		hf_fs.glob(f'{_REPOSITORY}/*/model.onnx')
		]
		_MODEL_NAMES = _open_models_for_repo_id(_REPO_ID).model_names


		class AnimePortraitBenchmark(BaseBenchmark):
		@@ -21,12 +14,10 @@ class AnimePortraitBenchmark(BaseBenchmark):
		self.model = model

		def load(self):
		from imgutils.validate.portrait import _open_anime_portrait_model
		_ = _open_anime_portrait_model(self.model)
		_open_models_for_repo_id(_REPO_ID)._open_model(self.model)

		def unload(self):
		from imgutils.validate.portrait import _open_anime_portrait_model
		_open_anime_portrait_model.cache_clear()
		_open_models_for_repo_id(_REPO_ID).clear()

		def run(self):
		image_file = random.choice(self.all_images)

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

deleted100644 → 0

+0 −2418

File deleted.

Preview size limit exceeded, changes collapsed.

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

+5 −14

Original line number	Diff line number	Diff line
		import os
		import random

		from huggingface_hub import HfFileSystem

		from benchmark import BaseBenchmark, create_plot_cli
		from imgutils.generic.classify import _open_models_for_repo_id
		from imgutils.validate import anime_style_age
		from imgutils.validate.style_age import _REPO_ID

		hf_fs = HfFileSystem()

		_REPOSITORY = 'deepghs/anime_style_ages'
		_MODEL_NAMES = [
		os.path.relpath(file, _REPOSITORY).split('/')[0] for file in
		hf_fs.glob(f'{_REPOSITORY}/*/model.onnx')
		]
		_MODEL_NAMES = _open_models_for_repo_id(_REPO_ID).model_names


		class AnimeStyleAgeBenchmark(BaseBenchmark):
		@@ -21,12 +14,10 @@ class AnimeStyleAgeBenchmark(BaseBenchmark):
		self.model = model

		def load(self):
		from imgutils.validate.style_age import _open_anime_style_age_model
		_ = _open_anime_style_age_model(self.model)
		_open_models_for_repo_id(_REPO_ID)._open_model(self.model)

		def unload(self):
		from imgutils.validate.style_age import _open_anime_style_age_model
		_open_anime_style_age_model.cache_clear()
		_open_models_for_repo_id(_REPO_ID).clear()

		def run(self):
		image_file = random.choice(self.all_images)

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

deleted100644 → 0

+0 −2274

File deleted.

Preview size limit exceeded, changes collapsed.

imgutils/validate/portrait.py

+6 −90

Original line number	Diff line number	Diff line
		@@ -15,16 +15,10 @@ Overview:
		The models are hosted on
		`huggingface - deepghs/anime_portrait <https://huggingface.co/deepghs/anime_portrait>`_.
		"""
		import json
		from functools import lru_cache
		from typing import Tuple, Optional, Dict, List
		from typing import Tuple, Dict

		import numpy as np
		from PIL import Image
		from huggingface_hub import hf_hub_download

		from imgutils.data import rgb_encode, ImageTyping, load_image
		from imgutils.utils import open_onnx_model
		from ..data import ImageTyping
		from ..generic import classify_predict, classify_predict_score

		__all__ = [
		'anime_portrait_score',
		@@ -32,81 +26,7 @@ __all__ = [
		]

		_DEFAULT_MODEL_NAME = 'mobilenetv3_v0_dist'


		@lru_cache()
		def _open_anime_portrait_model(model_name):
		"""
		Open the anime portrait model.

		:param model_name: The model name.
		:type model_name: str
		:return: The ONNX model.
		"""
		return open_onnx_model(hf_hub_download(
		f'deepghs/anime_portrait',
		f'{model_name}/model.onnx',
		))


		@lru_cache()
		def _get_anime_portrait_labels(model_name) -> List[str]:
		"""
		Get the labels for the anime portrait model.

		:param model_name: The model name.
		:type model_name: str
		:return: The list of labels.
		:rtype: List[str]
		"""
		with open(hf_hub_download(
		f'deepghs/anime_portrait',
		f'{model_name}/meta.json',
		), 'r') as f:
		return json.load(f)['labels']


		def _img_encode(image: Image.Image, size: Tuple[int, int] = (384, 384),
		normalize: Optional[Tuple[float, float]] = (0.5, 0.5)):
		"""
		Encode the input image.

		:param image: The input image.
		:type image: Image.Image
		:param size: The desired size of the image.
		:type size: Tuple[int, int]
		:param normalize: Mean and standard deviation for normalization. Default is (0.5, 0.5).
		:type normalize: Optional[Tuple[float, float]]
		:return: The encoded image data.
		:rtype: np.ndarray
		"""
		image = image.resize(size, Image.BILINEAR)
		data = rgb_encode(image, order_='CHW')

		if normalize is not None:
		mean_, std_ = normalize
		mean = np.asarray([mean_]).reshape((-1, 1, 1))
		std = np.asarray([std_]).reshape((-1, 1, 1))
		data = (data - mean) / std

		return data.astype(np.float32)


		def _raw_anime_portrait(image: ImageTyping, model_name: str = _DEFAULT_MODEL_NAME):
		"""
		Perform raw anime portrait processing on the input image.

		:param image: The input image.
		:type image: ImageTyping
		:param model_name: The model name. Default is 'mobilenetv3_v0_dist'.
		:type model_name: str
		:return: The processed image data.
		:rtype: np.ndarray
		"""
		image = load_image(image, force_background='white', mode='RGB')
		input_ = _img_encode(image)[None, ...]
		output, = _open_anime_portrait_model(model_name).run(['output'], {'input': input_})
		return output
		_REPO_ID = 'deepghs/anime_portrait'


		def anime_portrait_score(image: ImageTyping, model_name: str = _DEFAULT_MODEL_NAME) -> Dict[str, float]:
		@@ -142,9 +62,7 @@ def anime_portrait_score(image: ImageTyping, model_name: str = _DEFAULT_MODEL_NA
		>>> anime_portrait_score('head/9.jpg')
		{'person': 5.736660568800289e-07, 'halfbody': 7.199210472208506e-08, 'head': 0.9999992847442627}
		"""
		output = _raw_anime_portrait(image, model_name)
		values = dict(zip(_get_anime_portrait_labels(model_name), map(lambda x: x.item(), output[0])))
		return values
		return classify_predict_score(image, _REPO_ID, model_name)


		def anime_portrait(image: ImageTyping, model_name: str = _DEFAULT_MODEL_NAME) -> Tuple[str, float]:
		@@ -180,6 +98,4 @@ def anime_portrait(image: ImageTyping, model_name: str = _DEFAULT_MODEL_NAME) ->
		>>> anime_portrait('head/9.jpg')
		('head', 0.9999992847442627)
		"""
		output = _raw_anime_portrait(image, model_name)[0]
		max_id = np.argmax(output)
		return _get_anime_portrait_labels(model_name)[max_id], output[max_id].item()
		return classify_predict(image, _REPO_ID, model_name)