Update docs & OCR examples

README.md

@@ -40,13 +40,20 @@ The goal of Paddle Serving is to provide high-performance, flexible and easy-to-
 - Support service monitoring, provide prometheus-based performance statistics and port access
 
 
-<h2 align="center">Tutorial</h2>
+<h2 align="center">Tutorial and Papers</h2>
 
 
 - AIStudio tutorial(Chinese) : [Paddle Serving服务化部署框架](https://www.paddlepaddle.org.cn/tutorials/projectdetail/2538249)
+- AIStudio OCR practice(Chinese) : [基于PaddleServing的OCR服务化部署实战](https://aistudio.baidu.com/aistudio/projectdetail/3630726)
 - Video tutorial(Chinese) : [深度学习服务化部署-以互联网应用为例](https://aistudio.baidu.com/aistudio/course/introduce/19084)
 - Edge AI solution(Chinese) : [基于Paddle Serving&百度智能边缘BIE的边缘AI解决方案](https://mp.weixin.qq.com/s/j0EVlQXaZ7qmoz9Fv96Yrw)
 
+- Paper : [JiZhi: A Fast and Cost-Effective Model-As-A-Service System for
+Web-Scale Online Inference at Baidu](https://arxiv.org/pdf/2106.01674.pdf)
+- Paper : [ERNIE 3.0 TITAN: EXPLORING LARGER-SCALE KNOWLEDGE
+ENHANCED PRE-TRAINING FOR LANGUAGE UNDERSTANDING
+AND GENERATION](https://arxiv.org/pdf/2112.12731.pdf)
+
 <p align="center">
     <img src="doc/images/demo.gif" width="700">
 </p>
@@ -90,8 +97,6 @@ The first step is to call the model save interface to generate a model parameter
   - [Analyze and optimize performance](doc/Python_Pipeline/Performance_Tuning_EN.md)
   - [TensorRT dynamic Shape](doc/TensorRT_Dynamic_Shape_EN.md)
   - [Benchmark(Chinese)](doc/Python_Pipeline/Benchmark_CN.md)
-  - Our Paper: [JiZhi: A Fast and Cost-Effective Model-As-A-Service System for
-Web-Scale Online Inference at Baidu](https://arxiv.org/pdf/2106.01674.pdf)
 - Client SDK
   - [Python SDK(Chinese)](doc/C++_Serving/Introduction_CN.md#42-多语言多协议Client)
   - [JAVA SDK](doc/Java_SDK_EN.md)

README_CN.md

@@ -39,11 +39,18 @@ Paddle Serving依托深度学习框架PaddlePaddle旨在帮助深度学习开发
 - 支持服务监控，提供基于普罗米修斯的性能数据统计及端口访问
 
 
-<h2 align="center">教程</h2>
-
-- AIStudio教程-[Paddle Serving服务化部署框架](https://www.paddlepaddle.org.cn/tutorials/projectdetail/2538249)
-- 视频教程-[深度学习服务化部署-以互联网应用为例](https://aistudio.baidu.com/aistudio/course/introduce/19084)
-- 边缘AI解决方案-[基于Paddle Serving&百度智能边缘BIE的边缘AI解决方案](https://mp.weixin.qq.com/s/j0EVlQXaZ7qmoz9Fv96Yrw)
+<h2 align="center">教程与论文</h2>
+
+- AIStudio 使用教程 : [Paddle Serving服务化部署框架](https://www.paddlepaddle.org.cn/tutorials/projectdetail/2538249)
+- AIStudio OCR实战 : [基于PaddleServing的OCR服务化部署实战](https://aistudio.baidu.com/aistudio/projectdetail/3630726)
+- 视频教程 : [深度学习服务化部署-以互联网应用为例](https://aistudio.baidu.com/aistudio/course/introduce/19084)
+- 边缘AI 解决方案 : [基于Paddle Serving&百度智能边缘BIE的边缘AI解决方案](https://mp.weixin.qq.com/s/j0EVlQXaZ7qmoz9Fv96Yrw)
+
+- 论文 : [JiZhi: A Fast and Cost-Effective Model-As-A-Service System for
+Web-Scale Online Inference at Baidu](https://arxiv.org/pdf/2106.01674.pdf)
+- 论文 : [ERNIE 3.0 TITAN: EXPLORING LARGER-SCALE KNOWLEDGE
+ENHANCED PRE-TRAINING FOR LANGUAGE UNDERSTANDING
+AND GENERATION](https://arxiv.org/pdf/2112.12731.pdf)
 
 <p align="center">
     <img src="doc/images/demo.gif" width="700">

doc/Save_CN.md

@@ -2,6 +2,12 @@
 
 (简体中文|[English](./Save_EN.md))
 
+## 保存用于 Serving 部署模型的意义
+
+
+
+
+
 ## 从已保存的模型文件中导出
 如果已使用Paddle 的`save_inference_model`接口保存出预测要使用的模型，你可以使用Paddle Serving提供的名为`paddle_serving_client.convert`的内置模块进行转换。
 ```python

examples/Pipeline/PaddleOCR/ocr/pipeline_http_client.py

@@ -29,9 +29,26 @@ test_img_dir = "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()
+    # print file name
+    print('{}{}{}'.format('*' * 10, img_file, '*' * 10))
     image = cv2_to_base64(image_data1)
 
-for i in range(4):
     data = {"key": ["image"], "value": [image]}
     r = requests.post(url=url, data=json.dumps(data))
-    print(r.json())
+    result = r.json()
+    print("erro_no:{}, err_msg:{}".format(result["err_no"], result["err_msg"]))
+    # check success
+    if result["err_no"] == 0:
+        ocr_result = result["value"][0]
+        try:
+            for item in eval(ocr_result):
+                # return transcription and points
+                print("{}, {}".format(item[0], item[1]))
+        except Exception as e:
+            print("No results")
+            continue
+
+    else:
+        print(
+            "For details about error message, see PipelineServingLogs/pipeline.log.wf"
+        )

examples/Pipeline/PaddleOCR/ocr/pipeline_rpc_client.py

@@ -34,8 +34,24 @@ test_img_dir = "imgs/"
 for img_file in os.listdir(test_img_dir):
     with open(os.path.join(test_img_dir, img_file), 'rb') as file:
         image_data = file.read()
+    # print file name
+    print('{}{}{}'.format('*' * 10, img_file, '*' * 10))
     image = cv2_to_base64(image_data)
 
-for i in range(1):
-    ret = client.predict(feed_dict={"image": image}, fetch=["res"])
-    print(ret)
+    result = client.predict(feed_dict={"image": image}, fetch=["res"])
+    print("erro_no:{}, err_msg:{}".format(result.err_no, result.err_msg))
+    # check success
+    if result.err_no == 0:
+        ocr_result = result.value[0]
+        try:
+            for item in eval(ocr_result):
+                # return transcription and points
+                print("{}, {}".format(item[0], item[1]))
+        except Exception as e:
+            print("No results")
+            continue
+
+    else:
+        print(
+            "For details about error message, see PipelineServingLogs/pipeline.log.wf"
+        )

examples/Pipeline/PaddleOCR/ocr/web_service.py

@@ -14,6 +14,7 @@
 from paddle_serving_server.web_service import WebService, Op
 import logging
 import numpy as np
+import copy
 import cv2
 import base64
 from paddle_serving_app.reader import OCRReader
@@ -34,17 +35,18 @@ class DetOp(Op):
         self.filter_func = FilterBoxes(10, 10)
         self.post_func = DBPostProcess({
             "thresh": 0.3,
-            "box_thresh": 0.5,
+            "box_thresh": 0.6,
             "max_candidates": 1000,
             "unclip_ratio": 1.5,
             "min_size": 3
         })
-   
+
     """ 
     when opening tensorrt(configure in config.yml) and each time the input shape 
     for inferring is different, using this method for configuring tensorrt 
     dynamic shape to infer in each op model
     """
+
     def set_dynamic_shape_info(self):
         min_input_shape = {
             "x": [1, 3, 50, 50],
@@ -74,7 +76,7 @@ class DetOp(Op):
             "min_input_shape": min_input_shape,
             "max_input_shape": max_input_shape,
             "opt_input_shape": opt_input_shape,
-        }    
+        }
 
     def preprocess(self, input_dicts, data_id, log_id):
         (_, input_dict), = input_dicts.items()
@@ -107,25 +109,20 @@ class RecOp(Op):
         self.ocr_reader = OCRReader()
         self.get_rotate_crop_image = GetRotateCropImage()
         self.sorted_boxes = SortedBoxes()
-    
+
     """ 
     when opening tensorrt(configure in config.yml) and each time the input shape 
     for inferring is different, using this method for configuring tensorrt 
     dynamic shape to infer in each op model
     """
+
     def set_dynamic_shape_info(self):
-        min_input_shape = {
-            "x": [1, 3, 32, 10],
-            "lstm_1.tmp_0": [1, 1, 128]
-        }
+        min_input_shape = {"x": [1, 3, 32, 10], "lstm_1.tmp_0": [1, 1, 128]}
         max_input_shape = {
             "x": [50, 3, 32, 1000],
             "lstm_1.tmp_0": [500, 50, 128]
         }
-        opt_input_shape = {
-            "x": [6, 3, 32, 100],
-            "lstm_1.tmp_0": [25, 5, 128]
-        }
+        opt_input_shape = {"x": [6, 3, 32, 100], "lstm_1.tmp_0": [25, 5, 128]}
         self.dynamic_shape_info = {
             "min_input_shape": min_input_shape,
             "max_input_shape": max_input_shape,
@@ -137,8 +134,10 @@ class RecOp(Op):
         raw_im = input_dict["image"]
         data = np.frombuffer(raw_im, np.uint8)
         im = cv2.imdecode(data, cv2.IMREAD_COLOR)
-        dt_boxes = input_dict["dt_boxes"]
-        dt_boxes = self.sorted_boxes(dt_boxes)
+        self.dt_list = input_dict["dt_boxes"]
+        self.dt_list = self.sorted_boxes(self.dt_list)
+        # deepcopy to save origin dt_boxes
+        dt_boxes = copy.deepcopy(self.dt_list)
         feed_list = []
         img_list = []
         max_wh_ratio = 0
@@ -205,26 +204,31 @@ class RecOp(Op):
                 imgs[id] = norm_img
             feed = {"x": imgs.copy()}
             feed_list.append(feed)
-        #_LOGGER.info("feed_list : {}".format(feed_list))
 
         return feed_list, False, None, ""
 
     def postprocess(self, input_dicts, fetch_data, data_id, log_id):
-        res_list = []
+        rec_list = []
+        dt_num = len(self.dt_list)
         if isinstance(fetch_data, dict):
             if len(fetch_data) > 0:
                 rec_batch_res = self.ocr_reader.postprocess_ocrv2(
                     fetch_data, with_score=True)
                 for res in rec_batch_res:
-                    res_list.append(res[0])
+                    rec_list.append(res)
         elif isinstance(fetch_data, list):
             for one_batch in fetch_data:
                 one_batch_res = self.ocr_reader.postprocess_ocrv2(
                     one_batch, with_score=True)
                 for res in one_batch_res:
-                    res_list.append(res[0])
-
-        res = {"res": str(res_list)}
+                    rec_list.append(res)
+
+        result_list = []
+        for i in range(dt_num):
+            text = rec_list[i]
+            dt_box = self.dt_list[i]
+            result_list.append([text, dt_box.tolist()])
+        res = {"result": str(result_list)}
         return res, None, ""