Merge pull request #4 from PaddlePaddle/develop

update-2020-7-24

deploy/pdserving/ocr_web_client.py

+# Copyright (c) 2020 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.
+# -*- coding: utf-8 -*-
+
+import requests
+import json
+import cv2
+import base64
+import os, sys
+import time
+
+def cv2_to_base64(image):
+    #data = cv2.imencode('.jpg', image)[1]
+    return base64.b64encode(image).decode(
+        'utf8')  #data.tostring()).decode('utf8')
+
+headers = {"Content-type": "application/json"}
+url = "http://127.0.0.1:9292/ocr/prediction"
+test_img_dir = "../../doc/imgs/"
+for img_file in os.listdir(test_img_dir):
+    with open(os.path.join(test_img_dir, img_file), 'rb') as file:
+        image_data1 = file.read()
+    image = cv2_to_base64(image_data1)
+    data = {"feed": [{"image": image}], "fetch": ["res"]}
+    r = requests.post(url=url, headers=headers, data=json.dumps(data))
+    print(r.json())

deploy/pdserving/ocr_web_server.py

+# Copyright (c) 2020 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.
+
+from paddle_serving_client import Client
+from paddle_serving_app.reader import OCRReader
+import cv2
+import sys
+import numpy as np
+import os
+from paddle_serving_client import Client
+from paddle_serving_app.reader import Sequential, URL2Image, ResizeByFactor
+from paddle_serving_app.reader import Div, Normalize, Transpose
+from paddle_serving_app.reader import DBPostProcess, FilterBoxes, GetRotateCropImage, SortedBoxes
+from paddle_serving_server_gpu.web_service import WebService
+import time
+import re
+import base64
+
+
+class OCRService(WebService):
+    def init_det_client(self, det_port, det_client_config):
+        self.det_preprocess = Sequential([
+            ResizeByFactor(32, 960), Div(255),
+            Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225]), Transpose(
+                (2, 0, 1))
+        ])
+        self.det_client = Client()
+        self.det_client.load_client_config(det_client_config)
+        self.det_client.connect(["127.0.0.1:{}".format(det_port)])
+        self.ocr_reader = OCRReader()
+
+    def preprocess(self, feed=[], fetch=[]):
+        data = base64.b64decode(feed[0]["image"].encode('utf8'))
+        data = np.fromstring(data, np.uint8)
+        im = cv2.imdecode(data, cv2.IMREAD_COLOR)
+        ori_h, ori_w, _ = im.shape
+        det_img = self.det_preprocess(im)
+        det_out = self.det_client.predict(
+            feed={"image": det_img}, fetch=["concat_1.tmp_0"])
+        _, new_h, new_w = det_img.shape
+        filter_func = FilterBoxes(10, 10)
+        post_func = DBPostProcess({
+            "thresh": 0.3,
+            "box_thresh": 0.5,
+            "max_candidates": 1000,
+            "unclip_ratio": 1.5,
+            "min_size": 3
+        })
+        sorted_boxes = SortedBoxes()
+        ratio_list = [float(new_h) / ori_h, float(new_w) / ori_w]
+        dt_boxes_list = post_func(det_out["concat_1.tmp_0"], [ratio_list])
+        dt_boxes = filter_func(dt_boxes_list[0], [ori_h, ori_w])
+        dt_boxes = sorted_boxes(dt_boxes)
+        get_rotate_crop_image = GetRotateCropImage()
+        feed_list = []
+        img_list = []
+        max_wh_ratio = 0
+        for i, dtbox in enumerate(dt_boxes):
+            boximg = get_rotate_crop_image(im, dt_boxes[i])
+            img_list.append(boximg)
+            h, w = boximg.shape[0:2]
+            wh_ratio = w * 1.0 / h
+            max_wh_ratio = max(max_wh_ratio, wh_ratio)
+        for img in img_list:
+            norm_img = self.ocr_reader.resize_norm_img(img, max_wh_ratio)
+            feed = {"image": norm_img}
+            feed_list.append(feed)
+        fetch = ["ctc_greedy_decoder_0.tmp_0", "softmax_0.tmp_0"]
+        return feed_list, fetch
+
+    def postprocess(self, feed={}, fetch=[], fetch_map=None):
+        rec_res = self.ocr_reader.postprocess(fetch_map, with_score=True)
+        res_lst = []
+        for res in rec_res:
+            res_lst.append(res[0])
+        res = {"res": res_lst}
+        return res
+
+
+ocr_service = OCRService(name="ocr")
+ocr_service.load_model_config("ocr_rec_model")
+ocr_service.set_gpus("0")
+ocr_service.prepare_server(workdir="workdir", port=9292, device="gpu", gpuid=0)
+ocr_service.init_det_client(
+    det_port=9293,
+    det_client_config="ocr_det_client/serving_client_conf.prototxt")
+ocr_service.run_rpc_service()
+ocr_service.run_web_service()

deploy/pdserving/readme.md

-# Paddle Serving 服务部署
+# Paddle Serving 服务部署(Beta)
 
 本教程将介绍基于[Paddle Serving](https://github.com/PaddlePaddle/Serving)部署PaddleOCR在线预测服务的详细步骤。
 
 ## 快速启动服务
 
 ### 1. 准备环境
+我们先安装Paddle Serving相关组件
+我们推荐用户使用GPU来做Paddle Serving的OCR服务部署 
+
+**CUDA版本：9.0**
+**CUDNN版本：7.0**
+**操作系统版本：CentOS 6以上**
+
+```
+#以下提供beta版本的paddle serving whl包，欢迎试用，正式版会在7月底正式上线
+wget --no-check-certificate https://paddle-serving.bj.bcebos.com/others/paddle_serving_server_gpu-0.3.2-py2-none-any.whl
+wget --no-check-certificate https://paddle-serving.bj.bcebos.com/others/paddle_serving_app-0.1.2-py2-none-any.whl
+wget --no-check-certificate https://paddle-serving.bj.bcebos.com/others/paddle_serving_client-0.3.2-cp27-none-any.whl
+python -m pip install paddle_serving_app-0.1.2-py2-none-any.whl paddle_serving_server_gpu-0.3.2-py2-none-any.whl paddle_serving_client-0.3.2-cp27-none-any.whl
+```
 
 ### 2. 模型转换
+可以使用`paddle_serving_app`提供的模型，执行下列命令
+```
+python -m paddle_serving_app.package --get_model ocr_rec
+tar -xzvf ocr_rec.tar.gz
+python -m paddle_serving_app.package --get_model ocr_det
+tar -xzvf ocr_det.tar.gz 
+```
+执行上述命令会下载`db_crnn_mobile`的模型，如果想要下载规模更大的`db_crnn_server`模型，可以在下载预测模型并解压之后。参考[如何从Paddle保存的预测模型转为Paddle Serving格式可部署的模型](https://github.com/PaddlePaddle/Serving/blob/develop/doc/INFERENCE_TO_SERVING_CN.md)。
 
 ### 3. 启动服务
 启动服务可以根据实际需求选择启动`标准版`或者`快速版`，两种方式的对比如下表：  
 
 |版本|特点|适用场景|
 |-|-|-|
-|标准版|||
-|快速版|||
+|标准版|稳定性高，分布式部署|适用于吞吐量大，需要跨机房部署的情况|
+|快速版|部署方便，预测速度快|适用于对预测速度要求高，迭代速度快的场景|
 
 #### 方式1. 启动标准版服务
 
+```
+python -m paddle_serving_server_gpu.serve --model ocr_det_model --port 9293 --gpu_id 0
+python ocr_web_server.py
+```
+
 #### 方式2. 启动快速版服务
 
+```
+python ocr_local_server.py
+```
 
 ## 发送预测请求
 
+```
+python ocr_web_client.py
+```
+
 ## 返回结果格式说明
 
+返回结果是json格式
+```
+{u'result': {u'res': [u'\u571f\u5730\u6574\u6cbb\u4e0e\u571f\u58e4\u4fee\u590d\u7814\u7a76\u4e2d\u5fc3', u'\u534e\u5357\u519c\u4e1a\u5927\u5b661\u7d20\u56fe']}}
+```
+我们也可以打印结果json串中`res`字段的每一句话
+```
+土地整治与土壤修复研究中心
+华南农业大学1素图
+```
+
 ## 自定义修改服务逻辑
+
+在`ocr_web_server.py`或是`ocr_local_server.py`当中的`preprocess`函数里面做了检测服务和识别服务的前处理，`postprocess`函数里面做了识别的后处理服务，可以在相应的函数中做修改。调用了`paddle_serving_app`库提供的常见CV模型的前处理/后处理库。
+
+如果想要单独启动Paddle Serving的检测服务和识别服务，参见下列表格, 执行对应的脚本即可。
+
+| 模型 | 标准版         | 快速版           |
+| ---- | ----------------- | ------------------- |
+| 检测 | det_web_server.py | det_local_server.py |
+| 识别 | rec_web_server.py | rec_local_server.py |
+
+更多信息参见[Paddle Serving](https://github.com/PaddlePaddle/Serving)

deploy/pdserving/rec_local_server.py

+# Copyright (c) 2020 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.
+
+from paddle_serving_client import Client
+from paddle_serving_app.reader import OCRReader
+import cv2
+import sys
+import numpy as np
+import os
+from paddle_serving_client import Client
+from paddle_serving_app.reader import Sequential, URL2Image, ResizeByFactor
+from paddle_serving_app.reader import Div, Normalize, Transpose
+from paddle_serving_app.reader import DBPostProcess, FilterBoxes, GetRotateCropImage, SortedBoxes
+from paddle_serving_server_gpu.web_service import WebService
+import time
+import re
+import base64
+
+
+class OCRService(WebService):
+    def init_rec(self):
+        self.ocr_reader = OCRReader()
+
+    def preprocess(self, feed=[], fetch=[]):
+        img_list = []
+        for feed_data in feed:
+            data = base64.b64decode(feed_data["image"].encode('utf8'))
+            data = np.fromstring(data, np.uint8)
+            im = cv2.imdecode(data, cv2.IMREAD_COLOR)
+            img_list.append(im)
+        max_wh_ratio = 0
+        for i, boximg in enumerate(img_list):
+            h, w = boximg.shape[0:2]
+            wh_ratio = w * 1.0 / h
+            max_wh_ratio = max(max_wh_ratio, wh_ratio)
+        _, w, h = self.ocr_reader.resize_norm_img(img_list[0],
+                                                  max_wh_ratio).shape
+        imgs = np.zeros((len(img_list), 3, w, h)).astype('float32')
+        for i, img in enumerate(img_list):
+            norm_img = self.ocr_reader.resize_norm_img(img, max_wh_ratio)
+            imgs[i] = norm_img
+        feed = {"image": imgs.copy()}
+        fetch = ["ctc_greedy_decoder_0.tmp_0", "softmax_0.tmp_0"]
+        return feed, fetch
+
+    def postprocess(self, feed={}, fetch=[], fetch_map=None):
+        rec_res = self.ocr_reader.postprocess(fetch_map, with_score=True)
+        res_lst = []
+        for res in rec_res:
+            res_lst.append(res[0])
+        res = {"res": res_lst}
+        return res
+
+
+ocr_service = OCRService(name="ocr")
+ocr_service.load_model_config("ocr_rec_model")
+ocr_service.set_gpus("0")
+ocr_service.init_rec()
+ocr_service.prepare_server(workdir="workdir", port=9292, device="gpu", gpuid=0)
+ocr_service.run_debugger_service()
+ocr_service.run_web_service()

deploy/pdserving/rec_web_server.py

+# Copyright (c) 2020 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.
+
+from paddle_serving_client import Client
+from paddle_serving_app.reader import OCRReader
+import cv2
+import sys
+import numpy as np
+import os
+from paddle_serving_client import Client
+from paddle_serving_app.reader import Sequential, URL2Image, ResizeByFactor
+from paddle_serving_app.reader import Div, Normalize, Transpose
+from paddle_serving_app.reader import DBPostProcess, FilterBoxes, GetRotateCropImage, SortedBoxes
+from paddle_serving_server_gpu.web_service import WebService
+import time
+import re
+import base64
+
+
+class OCRService(WebService):
+    def init_rec(self):
+        self.ocr_reader = OCRReader()
+
+    def preprocess(self, feed=[], fetch=[]):
+        # TODO: to handle batch rec images
+        img_list = []
+        for feed_data in feed:
+            data = base64.b64decode(feed_data["image"].encode('utf8'))
+            data = np.fromstring(data, np.uint8)
+            im = cv2.imdecode(data, cv2.IMREAD_COLOR)
+            img_list.append(im)
+        feed_list = []
+        max_wh_ratio = 0
+        for i, boximg in enumerate(img_list):
+            h, w = boximg.shape[0:2]
+            wh_ratio = w * 1.0 / h
+            max_wh_ratio = max(max_wh_ratio, wh_ratio)
+        for img in img_list:
+            norm_img = self.ocr_reader.resize_norm_img(img, max_wh_ratio)
+            feed = {"image": norm_img}
+            feed_list.append(feed)
+        fetch = ["ctc_greedy_decoder_0.tmp_0", "softmax_0.tmp_0"]
+        return feed_list, fetch
+
+    def postprocess(self, feed={}, fetch=[], fetch_map=None):
+        rec_res = self.ocr_reader.postprocess(fetch_map, with_score=True)
+        res_lst = []
+        for res in rec_res:
+            res_lst.append(res[0])
+        res = {"res": res_lst}
+        return res
+
+
+ocr_service = OCRService(name="ocr")
+ocr_service.load_model_config("ocr_rec_model")
+ocr_service.set_gpus("0")
+ocr_service.init_rec()
+ocr_service.prepare_server(workdir="workdir", port=9292, device="gpu", gpuid=0)
+ocr_service.run_rpc_service()
+ocr_service.run_web_service()

doc/doc_ch/installation.md

@@ -80,3 +80,6 @@ git clone https://gitee.com/paddlepaddle/PaddleOCR
 cd PaddleOCR
 pip3 install -r requirments.txt
 ```
+
+注意，windows环境下，建议从[这里](https://www.lfd.uci.edu/~gohlke/pythonlibs/#shapely)下载shapely安装包完成安装，
+直接通过pip安装的shapely库可能出现`[winRrror 126] 找不到指定模块的问题`。

doc/doc_ch/serving.md

@@ -69,7 +69,7 @@ $ hub serving start --modules [Module1==Version1, Module2==Version2, ...] \
 
 #### 方式2. 配置文件启动（支持CPU、GPU）
 **启动命令：**  
-```hub serving start --config/-c config.json```  
+```hub serving start -c config.json```  
 
 其中，`config.json`格式如下： 
 ```python

doc/doc_ch/update.md

 # 更新
+- 2020.7.23 发布7月21日B站直播课回放和PPT，PaddleOCR开源大礼包全面解读，[获取地址](https://aistudio.baidu.com/aistudio/course/introduce/1519)
 - 2020.7.15 添加基于EasyEdge和Paddle-Lite的移动端DEMO，支持iOS和Android系统
 - 2020.7.15 完善预测部署，添加基于C++预测引擎推理、服务化部署和端侧部署方案，以及超轻量级中文OCR模型预测耗时Benchmark
 - 2020.7.15 整理OCR相关数据集、常用数据标注以及合成工具

doc/doc_en/installation_en.md

@@ -82,3 +82,9 @@ git clone https://gitee.com/paddlepaddle/PaddleOCR
 cd PaddleOCR
 pip3 install -r requirments.txt
 ```
+
+If you getting this error `OSError: [WinError 126] The specified module could not be found` when you install shapely on windows.
+
+Please try to download Shapely whl file using [http://www.lfd.uci.edu/~gohlke/pythonlibs/#shapely](http://www.lfd.uci.edu/~gohlke/pythonlibs/#shapely).
+
+Reference: [Solve shapely installation on windows](https://stackoverflow.com/questions/44398265/install-shapely-oserror-winerror-126-the-specified-module-could-not-be-found)

doc/doc_en/update_en.md

 # RECENT UPDATES
-
+- 2020.7.23, Release the playback and PPT of live class on BiliBili station, PaddleOCR Introduction, [address](https://aistudio.baidu.com/aistudio/course/introduce/1519)
 - 2020.7.15, Add mobile App demo , support both iOS and  Android  ( based on easyedge and Paddle Lite)
 - 2020.7.15, Improve the  deployment ability, add the C + +  inference , serving deployment. In addtion, the benchmarks of the ultra-lightweight Chinese OCR model are provided.
 - 2020.7.15, Add several related datasets, data annotation and synthesis tools.

ppocr/modeling/backbones/rec_mobilenet_v3.py

@@ -78,7 +78,7 @@ class MobileNetV3():
 
         supported_scale = [0.35, 0.5, 0.75, 1.0, 1.25]
         assert self.scale in supported_scale, \
-            "supported scale are {} but input scale is {}".format(supported_scale, scale)
+            "supported scales are {} but input scale is {}".format(supported_scale, self.scale)
 
     def __call__(self, input):
         scale = self.scale