Merge pull request #1686 from rainyfly/add_seeinthedark_module

add see in the dark module

modules/image/image_processing/seeinthedark/README.md

+# seeinthedark
+
+|模型名称|seeinthedark|
+| :--- | :---: |
+|类别|图像 - 暗光增强|
+|网络|ConvNet|
+|数据集|SID dataset|
+|是否支持Fine-tuning|否|
+|模型大小|120MB|
+|最新更新日期|2021-11-02|
+|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/22424850/142962370-a957d7b3-8050-4f5a-8462-3d6e49facb33.png"  width = "450" height = "300" hspace='10'/>
+    <br />
+    输入图像
+    <br />
+    <img src="https://user-images.githubusercontent.com/22424850/142962460-4a1b31ef-0eec-423b-ab3d-8622f3e8261a.png"  width = "450" height = "300" hspace='10'/>
+    <br />
+    输出图像
+     <br />
+    </p>
+
+- ### 模型介绍
+
+  - 通过大量暗光条件下短曝光和长曝光组成的图像对，以RAW图像为输入，RGB图像为参照进行训练，该模型实现端到端直接将暗光下的RAW图像处理得到可见的RGB图像。
+
+  - 更多详情参考：[Learning to See in the Dark](http://cchen156.github.io/paper/18CVPR_SID.pdf)
+
+
+
+## 二、安装
+
+- ### 1、环境依赖  
+  - rawpy
+
+- ### 2、安装
+
+  - ```shell
+    $ hub install seeinthedark
+    ```
+  - 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 raw(Sony, .ARW) file
+    $ hub run seeinthedark --input_path "/PATH/TO/IMAGE"
+    ```
+  - 通过命令行方式实现暗光增强模型的调用，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
+
+    denoiser = hub.Module(name="seeinthedark")
+    input_path = "/PATH/TO/IMAGE"
+    # Read from a raw file
+    denoiser.denoising(paths=[input_path], output_path='./denoising_result.png', use_gpu=True)  
+    ```
+
+- ### 3、API
+
+  - ```python
+    def denoising(images=None, paths=None, output_dir='./denoising_result/', use_gpu=False, visualization=True)
+    ```
+    - 暗光增强API，完成对暗光RAW图像的降噪并处理生成RGB图像。
+
+    - **参数**
+      - images (list\[numpy.ndarray\]): 输入的图像，单通道的马赛克图像; <br/>
+      - paths (list\[str\]): 暗光图像文件的路径，Sony的RAW格式；<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可以部署一个在线图像风格转换服务。
+
+- ### 第一步：启动PaddleHub Serving
+
+  - 运行启动命令：
+  - ```shell
+    $ hub serving start -m seeinthedark
+    ```
+
+  - 这样就完成了一个图像风格转换的在线服务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 rawpy
+    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(rawpy.imread("/PATH/TO/IMAGE").raw_image_visible)]}
+    headers = {"Content-type": "application/json"}
+    url = "http://127.0.0.1:8866/predict/seeinthedark/"
+    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 seeinthedark==1.0.0
+    ```

modules/image/image_processing/seeinthedark/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 os
+import argparse
+
+import paddle
+import paddlehub as hub
+from paddlehub.module.module import moduleinfo, runnable, serving
+import numpy as np
+import rawpy
+import cv2
+
+from .util import base64_to_cv2
+
+
+def pack_raw(raw):
+    # pack Bayer image to 4 channels
+    im = raw
+    if not isinstance(raw, np.ndarray):
+        im = raw.raw_image_visible.astype(np.float32)
+    im = np.maximum(im - 512, 0) / (16383 - 512)  # subtract the black level
+
+    im = np.expand_dims(im, axis=2)
+    img_shape = im.shape
+    H = img_shape[0]
+    W = img_shape[1]
+
+    out = np.concatenate((im[0:H:2, 0:W:2, :], im[0:H:2, 1:W:2, :], im[1:H:2, 1:W:2, :], im[1:H:2, 0:W:2, :]), axis=2)
+    return out
+
+
+@moduleinfo(
+    name="seeinthedark", type="CV/denoising", author="paddlepaddle", author_email="", summary="", version="1.0.0")
+class LearningToSeeInDark:
+    def __init__(self):
+        self.pretrained_model = os.path.join(self.directory, "pd_model/inference_model")
+        self.cpu_have_loaded = False
+        self.gpu_have_loaded = False
+
+    def set_device(self, use_gpu=False):
+        if use_gpu == False:
+            if not self.cpu_have_loaded:
+                exe = paddle.static.Executor(paddle.CPUPlace())
+                [prog, inputs, outputs] = paddle.static.load_inference_model(
+                    path_prefix=self.pretrained_model,
+                    executor=exe,
+                    model_filename="model.pdmodel",
+                    params_filename="model.pdiparams")
+                self.cpuexec, self.cpuprog, self.cpuinputs, self.cpuoutputs = exe, prog, inputs, outputs
+                self.cpu_have_loaded = True
+
+            return self.cpuexec, self.cpuprog, self.cpuinputs, self.cpuoutputs
+
+        else:
+            if not self.gpu_have_loaded:
+                exe = paddle.static.Executor(paddle.CUDAPlace(0))
+                [prog, inputs, outputs] = paddle.static.load_inference_model(
+                    path_prefix=self.pretrained_model,
+                    executor=exe,
+                    model_filename="model.pdmodel",
+                    params_filename="model.pdiparams")
+                self.gpuexec, self.gpuprog, self.gpuinputs, self.gpuoutputs = exe, prog, inputs, outputs
+                self.gpu_have_loaded = True
+
+            return self.gpuexec, self.gpuprog, self.gpuinputs, self.gpuoutputs
+
+    def denoising(self,
+                  images: list = None,
+                  paths: list = None,
+                  output_dir: str = './enlightening_result/',
+                  use_gpu: bool = False,
+                  visualization: bool = True):
+        '''
+        Denoise a raw image in the low-light scene.
+
+        images (list[numpy.ndarray]): data of images, shape of each is [H, W], must be sing-channel image captured by camera.
+        paths (list[str]): paths to images
+        output_dir: the dir to save the results
+        use_gpu: if True, use gpu to perform the computation, otherwise cpu.
+        visualization: if True, save results in output_dir.
+        '''
+        results = []
+        paddle.enable_static()
+        exe, prog, inputs, outputs = self.set_device(use_gpu)
+
+        if images != None:
+            for raw in images:
+                input_full = np.expand_dims(pack_raw(raw), axis=0) * 300
+                px = input_full.shape[1] // 512
+                py = input_full.shape[2] // 512
+                rx, ry = px * 512, py * 512
+                input_full = input_full[:, :rx, :ry, :]
+                output = np.random.randn(rx * 2, ry * 2, 3)
+                input_full = np.minimum(input_full, 1.0)
+                for i in range(px):
+                    for j in range(py):
+                        input_patch = input_full[:, i * 512:i * 512 + 512, j * 512:j * 512 + 512, :]
+                        result = exe.run(prog, feed={inputs[0]: input_patch}, fetch_list=outputs)
+                        output[i * 512 * 2:i * 512 * 2 + 512 * 2, j * 512 * 2:j * 512 * 2 + 512 * 2, :] = result[0][0]
+                output = np.minimum(np.maximum(output, 0), 1)
+                output = output * 255
+                output = np.clip(output, 0, 255)
+                output = output.astype('uint8')
+                results.append(output)
+        if paths != None:
+            for path in paths:
+                raw = rawpy.imread(path)
+                input_full = np.expand_dims(pack_raw(raw), axis=0) * 300
+                px = input_full.shape[1] // 512
+                py = input_full.shape[2] // 512
+                rx, ry = px * 512, py * 512
+                input_full = input_full[:, :rx, :ry, :]
+                output = np.random.randn(rx * 2, ry * 2, 3)
+                input_full = np.minimum(input_full, 1.0)
+                for i in range(px):
+                    for j in range(py):
+                        input_patch = input_full[:, i * 512:i * 512 + 512, j * 512:j * 512 + 512, :]
+                        result = exe.run(prog, feed={inputs[0]: input_patch}, fetch_list=outputs)
+                        output[i * 512 * 2:i * 512 * 2 + 512 * 2, j * 512 * 2:j * 512 * 2 + 512 * 2, :] = result[0][0]
+                output = np.minimum(np.maximum(output, 0), 1)
+                output = output * 255
+                output = np.clip(output, 0, 255)
+                output = output.astype('uint8')
+                results.append(output)
+
+        if visualization == True:
+            if not os.path.exists(output_dir):
+                os.makedirs(output_dir, exist_ok=True)
+            for i, out in enumerate(results):
+                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)
+        self.denoising(
+            paths=[self.args.input_path],
+            output_dir=self.args.output_dir,
+            use_gpu=self.args.use_gpu,
+            visualization=self.args.visualization)
+
+    @serving
+    def serving_method(self, images, **kwargs):
+        """
+        Run as a service.
+        """
+        images_decode = [base64_to_cv2(image) for image in images]
+        results = self.denoising(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='denoising_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 raw image, should be raw file captured by camera.")

modules/image/image_processing/seeinthedark/requirements.txt

+rawpy