add swinir_l_real_sr_x4 (#2076)

* git add swinir_l_real_sr_x4

* fix typo

* fix typo
Co-authored-by: Nchenjian <chenjian26@baidu.com>

modules/image/Image_editing/super_resolution/swinir_l_real_sr_x4/README.md

+# swinir_l_real_sr_x4
+
+|模型名称|swinir_l_real_sr_x4|
+| :--- | :---: |
+|类别|图像-图像编辑|
+|网络|SwinIR|
+|数据集|DIV2K / Flickr2K|
+|是否支持Fine-tuning|否|
+|模型大小|142.2MB|
+|指标|-|
+|最新更新日期|2022-10-10|
+
+
+## 一、模型基本信息
+
+- ### 应用效果展示
+
+  - 网络结构：
+      <p align="center">
+      <img src="https://ai-studio-static-online.cdn.bcebos.com/b3c6bfc3dfc14078adcf3dc19acaf04acd4b064770384e2bbd8865697c7dbc91" hspace='10'/> <br />
+      </p>
+
+  - 样例结果示例：
+      <p align="center">
+      <img src="https://ai-studio-static-online.cdn.bcebos.com/c5517af6c3f944c4b281aedc417a4f8c02c0a969d0dd494c9106c4ff2709fc2f" hspace='10'/>
+      <img src="https://ai-studio-static-online.cdn.bcebos.com/183c5821029f45bbb78d1700ab8297baabba15f82ab4467e88414bbed056ccf0" hspace='10'/>
+      </p>
+
+- ### 模型介绍
+
+  - SwinIR 是一个基于 Swin Transformer 的图像恢复模型。swinir_l_real_sr_x4 是基于 SwinIR-L 的 4 倍现实图像超分辨率模型。
+
+
+
+## 二、安装
+
+- ### 1、环境依赖
+
+  - paddlepaddle >= 2.0.0
+
+  - paddlehub >= 2.0.0  
+
+- ### 2.安装
+
+    - ```shell
+      $ hub install swinir_l_real_sr_x4
+      ```
+    -  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
+    $ hub run swinir_l_real_sr_x4 \
+        --input_path "/PATH/TO/IMAGE" \
+        --output_dir "swinir_l_real_sr_x4_output"
+    ```
+
+  - ### 2、预测代码示例
+
+    ```python
+    import paddlehub as hub
+    import cv2
+
+    module = hub.Module(name="swinir_l_real_sr_x4")
+    result = module.real_sr(
+        image=cv2.imread('/PATH/TO/IMAGE'),
+        visualization=True,
+        output_dir='swinir_l_real_sr_x4_output'
+    )
+    ```
+
+  - ### 3、API
+
+    ```python
+    def real_sr(
+        image: Union[str, numpy.ndarray],
+        visualization: bool = True,
+        output_dir: str = "swinir_l_real_sr_x4_output"
+    ) -> numpy.ndarray
+    ```
+
+    - 超分辨率 API
+
+    - **参数**
+
+      * image (Union\[str, numpy.ndarray\]): 图片数据，ndarray.shape 为 \[H, W, C\]，BGR格式；
+      * visualization (bool): 是否将识别结果保存为图片文件；
+      * output\_dir (str): 保存处理结果的文件目录。
+
+    - **返回**
+
+      * res (numpy.ndarray): 图像超分辨率结果 (BGR)；
+
+## 四、服务部署
+
+- PaddleHub Serving 可以部署一个图像超分辨率的在线服务。
+
+- ### 第一步：启动PaddleHub Serving
+
+  - 运行启动命令：
+
+    ```shell
+     $ hub serving start -m swinir_l_real_sr_x4
+    ```
+
+    - 这样就完成了一个图像超分辨率服务化API的部署，默认端口号为8866。
+
+- ### 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 base64
+
+    import cv2
+    import numpy as np
+
+    def cv2_to_base64(image):
+        data = cv2.imencode('.jpg', image)[1]
+        return base64.b64encode(data.tobytes()).decode('utf8')
+
+    def base64_to_cv2(b64str):
+        data = base64.b64decode(b64str.encode('utf8'))
+        data = np.frombuffer(data, np.uint8)
+        data = cv2.imdecode(data, cv2.IMREAD_COLOR)
+        return data
+
+    # 发送HTTP请求
+    org_im = cv2.imread('/PATH/TO/IMAGE')
+    data = {
+        'image': cv2_to_base64(org_im)
+    }
+    headers = {"Content-type": "application/json"}
+    url = "http://127.0.0.1:8866/predict/swinir_l_real_sr_x4"
+    r = requests.post(url=url, headers=headers, data=json.dumps(data))
+
+    # 结果转换
+    results = r.json()['results']
+    results = base64_to_cv2(results)
+
+    # 保存结果
+    cv2.imwrite('output.jpg', results)
+    ```
+
+## 五、参考资料
+
+* 论文：[SwinIR: Image Restoration Using Swin Transformer](https://arxiv.org/abs/2108.10257)
+
+* 官方实现：[JingyunLiang/SwinIR](https://github.com/JingyunLiang/SwinIR)
+
+## 六、更新历史
+
+* 1.0.0
+
+  初始发布
+
+  ```shell
+  $ hub install swinir_l_real_sr_x4==1.0.0
+  ```

modules/image/Image_editing/super_resolution/swinir_l_real_sr_x4/module.py

+import argparse
+import base64
+import os
+import time
+from typing import Union
+
+import cv2
+import numpy as np
+import paddle
+import paddle.nn as nn
+
+from .swinir import SwinIR
+from paddlehub.module.module import moduleinfo
+from paddlehub.module.module import runnable
+from paddlehub.module.module import serving
+
+
+def cv2_to_base64(image):
+    data = cv2.imencode('.jpg', image)[1]
+    return base64.b64encode(data.tobytes()).decode('utf8')
+
+
+def base64_to_cv2(b64str):
+    data = base64.b64decode(b64str.encode('utf8'))
+    data = np.frombuffer(data, np.uint8)
+    data = cv2.imdecode(data, cv2.IMREAD_COLOR)
+    return data
+
+
+@moduleinfo(
+    name='swinir_l_real_sr_x4',
+    version='1.0.0',
+    type="CV/image_editing",
+    author="",
+    author_email="",
+    summary="Image Restoration (Real image Super Resolution) Using Swin Transformer.",
+)
+class SwinIRMRealSR(nn.Layer):
+
+    def __init__(self):
+        super(SwinIRMRealSR, self).__init__()
+        self.default_pretrained_model_path = os.path.join(self.directory,
+                                                          '003_realSR_BSRGAN_DFOWMFC_s64w8_SwinIR-L_x4_GAN.pdparams')
+        self.swinir = SwinIR(upscale=4,
+                             in_chans=3,
+                             img_size=64,
+                             window_size=8,
+                             img_range=1.,
+                             depths=[6, 6, 6, 6, 6, 6, 6, 6, 6],
+                             embed_dim=240,
+                             num_heads=[8, 8, 8, 8, 8, 8, 8, 8, 8],
+                             mlp_ratio=2,
+                             upsampler='nearest+conv',
+                             resi_connection='3conv')
+        state_dict = paddle.load(self.default_pretrained_model_path)
+        self.swinir.set_state_dict(state_dict)
+        self.swinir.eval()
+
+    def preprocess(self, img: np.ndarray) -> np.ndarray:
+        img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
+        img = img.transpose((2, 0, 1))
+        img = img / 255.0
+        return img.astype(np.float32)
+
+    def postprocess(self, img: np.ndarray) -> np.ndarray:
+        img = img.clip(0, 1)
+        img = img * 255.0
+        img = img.transpose((1, 2, 0))
+        img = cv2.cvtColor(img, cv2.COLOR_RGB2BGR)
+        return img.astype(np.uint8)
+
+    def real_sr(self,
+                image: Union[str, np.ndarray],
+                visualization: bool = True,
+                output_dir: str = "swinir_l_real_sr_x4_output") -> np.ndarray:
+        if isinstance(image, str):
+            _, file_name = os.path.split(image)
+            save_name, _ = os.path.splitext(file_name)
+            save_name = save_name + '_' + str(int(time.time())) + '.jpg'
+            image = cv2.imdecode(np.fromfile(image, dtype=np.uint8), cv2.IMREAD_COLOR)
+        elif isinstance(image, np.ndarray):
+            save_name = str(int(time.time())) + '.jpg'
+            image = image
+        else:
+            raise Exception("image should be a str / np.ndarray")
+
+        with paddle.no_grad():
+            img_input = self.preprocess(image)
+            img_input = paddle.to_tensor(img_input[None, ...], dtype=paddle.float32)
+
+            img_output = self.swinir(img_input)
+            img_output = img_output.numpy()[0]
+            img_output = self.postprocess(img_output)
+
+        if visualization:
+            if not os.path.isdir(output_dir):
+                os.makedirs(output_dir)
+            save_path = os.path.join(output_dir, save_name)
+            cv2.imwrite(save_path, img_output)
+
+        return img_output
+
+    @runnable
+    def run_cmd(self, argvs):
+        """
+        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.parser.add_argument('--input_path', type=str, help="Path to image.")
+        self.parser.add_argument('--output_dir',
+                                 type=str,
+                                 default='swinir_l_real_sr_x4_output',
+                                 help="The directory to save output images.")
+        args = self.parser.parse_args(argvs)
+        self.real_sr(image=args.input_path, visualization=True, output_dir=args.output_dir)
+        return 'Results are saved in %s' % args.output_dir
+
+    @serving
+    def serving_method(self, image, **kwargs):
+        """
+        Run as a service.
+        """
+        image = base64_to_cv2(image)
+        img_output = self.real_sr(image=image, **kwargs)
+
+        return cv2_to_base64(img_output)

modules/image/Image_editing/super_resolution/swinir_l_real_sr_x4/test.py

+import os
+import shutil
+import unittest
+
+import cv2
+import numpy as np
+import requests
+
+import paddlehub as hub
+
+os.environ['CUDA_VISIBLE_DEVICES'] = '0'
+
+
+class TestHubModule(unittest.TestCase):
+
+    @classmethod
+    def setUpClass(cls) -> None:
+        img_url = 'https://unsplash.com/photos/mJaD10XeD7w/download?ixid=MnwxMjA3fDB8MXxzZWFyY2h8M3x8Y2F0fGVufDB8fHx8MTY2MzczNDc3Mw&force=true&w=640'
+        if not os.path.exists('tests'):
+            os.makedirs('tests')
+        response = requests.get(img_url)
+        assert response.status_code == 200, 'Network Error.'
+        with open('tests/test.jpg', 'wb') as f:
+            f.write(response.content)
+        img = cv2.imread('tests/test.jpg')
+        img = cv2.resize(img, (0, 0), fx=0.25, fy=0.25)
+        cv2.imwrite('tests/test.jpg', img)
+        cls.module = hub.Module(name="swinir_l_real_sr_x4")
+
+    @classmethod
+    def tearDownClass(cls) -> None:
+        shutil.rmtree('tests')
+        shutil.rmtree('swinir_l_real_sr_x4_output')
+
+    def test_real_sr1(self):
+        results = self.module.real_sr(image='tests/test.jpg', visualization=False)
+
+        self.assertIsInstance(results, np.ndarray)
+
+    def test_real_sr2(self):
+        results = self.module.real_sr(image=cv2.imread('tests/test.jpg'), visualization=True)
+
+        self.assertIsInstance(results, np.ndarray)
+
+    def test_real_sr3(self):
+        results = self.module.real_sr(image=cv2.imread('tests/test.jpg'), visualization=True)
+
+        self.assertIsInstance(results, np.ndarray)
+
+    def test_real_sr4(self):
+        self.assertRaises(Exception, self.module.real_sr, image=['tests/test.jpg'])
+
+    def test_real_sr5(self):
+        self.assertRaises(FileNotFoundError, self.module.real_sr, image='no.jpg')
+
+
+if __name__ == "__main__":
+    unittest.main()