Merge pull request #1717 from rainyfly/face_parse

add face_parse module

modules/image/Image_gan/style_transfer/face_parse/README.md

+# face_parse
+
+|模型名称|face_parse|
+| :--- | :---: |
+|类别|图像 - 人脸解析|
+|网络|BiSeNet|
+|数据集|COCO-Stuff|
+|是否支持Fine-tuning|否|
+|模型大小|77MB|
+|最新更新日期|2021-12-07|
+|数据指标|-|
+
+
+## 一、模型基本信息  
+
+- ### 应用效果展示
+  - 样例结果示例：
+    <p align="center">
+    <img src="https://user-images.githubusercontent.com/22424850/157190651-595b6964-97c5-4b0b-ac0a-c30c8520a972.png"  width = "40%"  hspace='10'/>
+    <br />
+    输入图像
+    <br />
+    <img src="https://user-images.githubusercontent.com/22424850/157192693-b3f737ed-1a24-4ef9-8454-bfd9d51755af.png"  width = "40%"  hspace='10'/>
+    <br />
+    输出图像
+     <br />
+    </p>
+
+- ### 模型介绍
+
+  - 人脸解析是语义图像分割的一种特殊情况，人脸解析是计算人脸图像中不同语义成分(如头发、嘴唇、鼻子、眼睛等)的像素级标签映射。给定一个输入的人脸图像，人脸解析将为每个语义成分分配一个像素级标签。
+
+
+
+## 二、安装
+
+- ### 1、环境依赖  
+  - ppgan
+  - dlib
+
+- ### 2、安装
+
+  - ```shell
+    $ hub install face_parse
+    ```
+  - 如您安装时遇到问题，可参考：[零基础windows安装](../../../../docs/docs_ch/get_start/windows_quickstart.md)
+ | [零基础Linux安装](../../../../docs/docs_ch/get_start/linux_quickstart.md) | [零基础MacOS安装](../../../../docs/docs_ch/get_start/mac_quickstart.md)
+
+## 三、模型API预测
+
+- ### 1、命令行预测
+
+  - ```shell
+    # Read from a file
+    $ hub run face_parse --input_path "/PATH/TO/IMAGE"
+    ```
+  - 通过命令行方式实现人脸解析模型的调用，更多请见 [PaddleHub命令行指令](../../../../docs/docs_ch/tutorial/cmd_usage.rst)
+
+- ### 2、预测代码示例
+
+  - ```python
+    import paddlehub as hub
+
+    module = hub.Module(name="face_parse")
+    input_path = ["/PATH/TO/IMAGE"]
+    # Read from a file
+    module.style_transfer(paths=input_path, output_dir='./transfer_result/', use_gpu=True)  
+    ```
+
+- ### 3、API
+
+  - ```python
+    style_transfer(images=None, paths=None, output_dir='./transfer_result/', use_gpu=False, visualization=True):
+    ```
+    - 人脸解析转换API。
+
+    - **参数**
+
+      - images (list\[numpy.ndarray\]): 图片数据，ndarray.shape 为 \[H, W, C\]；<br/>
+      - paths (list\[str\]): 图片的路径；<br/>
+      - output\_dir (str): 结果保存的路径； <br/>
+      - use\_gpu (bool): 是否使用 GPU；<br/>
+      - visualization(bool): 是否保存结果到本地文件夹
+
+
+## 四、服务部署
+
+- PaddleHub Serving可以部署一个在线人脸解析转换服务。
+
+- ### 第一步：启动PaddleHub Serving
+
+  - 运行启动命令：
+  - ```shell
+    $ hub serving start -m face_parse
+    ```
+
+  - 这样就完成了一个人脸解析转换的在线服务API的部署，默认端口号为8866。
+
+  - **NOTE:** 如使用GPU预测，则需要在启动服务之前，请设置CUDA\_VISIBLE\_DEVICES环境变量，否则不用设置。
+
+- ### 第二步：发送预测请求
+
+  - 配置好服务端，以下数行代码即可实现发送预测请求，获取预测结果
+
+  - ```python
+    import requests
+    import json
+    import cv2
+    import base64
+
+
+    def cv2_to_base64(image):
+      data = cv2.imencode('.jpg', image)[1]
+      return base64.b64encode(data.tostring()).decode('utf8')
+
+    # 发送HTTP请求
+    data = {'images':[cv2_to_base64(cv2.imread("/PATH/TO/IMAGE"))]}
+    headers = {"Content-type": "application/json"}
+    url = "http://127.0.0.1:8866/predict/face_parse"
+    r = requests.post(url=url, headers=headers, data=json.dumps(data))
+
+    # 打印预测结果
+    print(r.json()["results"])
+
+## 五、更新历史
+
+* 1.0.0
+
+  初始发布
+
+  - ```shell
+    $ hub install face_parse==1.0.0
+    ```

modules/image/Image_gan/style_transfer/face_parse/model.py

+# copyright (c) 2020 PaddlePaddle Authors. All Rights Reserve.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import os
+import sys
+import argparse
+
+from PIL import Image
+import numpy as np
+import cv2
+
+import ppgan.faceutils as futils
+from ppgan.utils.preprocess import *
+from ppgan.utils.visual import mask2image
+
+
+class FaceParsePredictor:
+    def __init__(self):
+        self.input_size = (512, 512)
+        self.up_ratio = 0.6 / 0.85
+        self.down_ratio = 0.2 / 0.85
+        self.width_ratio = 0.2 / 0.85
+        self.face_parser = futils.mask.FaceParser()
+
+    def run(self, image):
+        image = Image.fromarray(image)
+        face = futils.dlib.detect(image)
+
+        if not face:
+            return
+        face_on_image = face[0]
+        image, face, crop_face = futils.dlib.crop(image, face_on_image, self.up_ratio, self.down_ratio,
+                                                  self.width_ratio)
+        np_image = np.array(image)
+        mask = self.face_parser.parse(np.float32(cv2.resize(np_image, self.input_size)))
+        mask = cv2.resize(mask.numpy(), (256, 256))
+        mask = mask.astype(np.uint8)
+        mask = mask2image(mask)
+
+        return mask

modules/image/Image_gan/style_transfer/face_parse/module.py

+# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+import argparse
+import copy
+import os
+
+import cv2
+import numpy as np
+import paddle
+from skimage.io import imread
+from skimage.transform import rescale
+from skimage.transform import resize
+
+import paddlehub as hub
+from .model import FaceParsePredictor
+from .util import base64_to_cv2
+from paddlehub.module.module import moduleinfo
+from paddlehub.module.module import runnable
+from paddlehub.module.module import serving
+
+
+@moduleinfo(
+    name="face_parse", type="CV/style_transfer", author="paddlepaddle", author_email="", summary="", version="1.0.0")
+class Face_parse:
+    def __init__(self):
+        self.pretrained_model = os.path.join(self.directory, "bisenet.pdparams")
+
+        self.network = FaceParsePredictor()
+
+    def style_transfer(self,
+                       images: list = None,
+                       paths: list = None,
+                       output_dir: str = './transfer_result/',
+                       use_gpu: bool = False,
+                       visualization: bool = True):
+        '''
+
+
+        images (list[numpy.ndarray]): data of images, shape of each is [H, W, C], color space must be BGR(read by cv2).
+        paths (list[str]): paths to images
+        output_dir (str): the dir to save the results
+        use_gpu (bool): if True, use gpu to perform the computation, otherwise cpu.
+        visualization (bool): if True, save results in output_dir.
+        '''
+        results = []
+        paddle.disable_static()
+        place = 'gpu:0' if use_gpu else 'cpu'
+        place = paddle.set_device(place)
+        if images == None and paths == None:
+            print('No image provided. Please input an image or a image path.')
+            return
+
+        if images != None:
+            for image in images:
+                image = image[:, :, ::-1]
+                out = self.network.run(image)
+                results.append(out)
+
+        if paths != None:
+            for path in paths:
+                image = cv2.imread(path)[:, :, ::-1]
+                out = self.network.run(image)
+                results.append(out)
+
+        if visualization == True:
+            if not os.path.exists(output_dir):
+                os.makedirs(output_dir, exist_ok=True)
+            for i, out in enumerate(results):
+                if out is not None:
+                    cv2.imwrite(os.path.join(output_dir, 'output_{}.png'.format(i)), out[:, :, ::-1])
+
+        return results
+
+    @runnable
+    def run_cmd(self, argvs: list):
+        """
+        Run as a command.
+        """
+        self.parser = argparse.ArgumentParser(
+            description="Run the {} module.".format(self.name),
+            prog='hub run {}'.format(self.name),
+            usage='%(prog)s',
+            add_help=True)
+
+        self.arg_input_group = self.parser.add_argument_group(title="Input options", description="Input data. Required")
+        self.arg_config_group = self.parser.add_argument_group(
+            title="Config options", description="Run configuration for controlling module behavior, not required.")
+        self.add_module_config_arg()
+        self.add_module_input_arg()
+        self.args = self.parser.parse_args(argvs)
+        results = self.style_transfer(
+            paths=[self.args.input_path],
+            output_dir=self.args.output_dir,
+            use_gpu=self.args.use_gpu,
+            visualization=self.args.visualization)
+        return results
+
+    @serving
+    def serving_method(self, images, **kwargs):
+        """
+        Run as a service.
+        """
+        images_decode = [base64_to_cv2(image) for image in images]
+        results = self.style_transfer(images=images_decode, **kwargs)
+        tolist = [result.tolist() for result in results]
+        return tolist
+
+    def add_module_config_arg(self):
+        """
+        Add the command config options.
+        """
+        self.arg_config_group.add_argument('--use_gpu', action='store_true', help="use GPU or not")
+
+        self.arg_config_group.add_argument(
+            '--output_dir', type=str, default='transfer_result', help='output directory for saving result.')
+        self.arg_config_group.add_argument('--visualization', type=bool, default=False, help='save results or not.')
+
+    def add_module_input_arg(self):
+        """
+        Add the command input options.
+        """
+        self.arg_input_group.add_argument('--input_path', type=str, help="path to input image.")

modules/image/Image_gan/style_transfer/face_parse/requirements.txt

+ppgan
+dlib

modules/image/Image_gan/style_transfer/face_parse/util.py

+import base64
+import cv2
+import numpy as np
+
+
+def base64_to_cv2(b64str):
+    data = base64.b64decode(b64str.encode('utf8'))
+    data = np.fromstring(data, np.uint8)
+    data = cv2.imdecode(data, cv2.IMREAD_COLOR)
+    return data