update chinese_ocr_db_crnn_server (#2171)

modules/image/text_recognition/chinese_ocr_db_crnn_server/README.md

@@ -64,7 +64,7 @@
     ```
   - 通过命令行方式实现文字识别模型的调用，更多请见 [PaddleHub命令行指令](../../../../docs/docs_ch/tutorial/cmd_usage.rst)
 
-- ### 2、代码示例
+- ### 2、预测代码示例
 
   - ```python
     import paddlehub as hub
@@ -165,6 +165,8 @@
     print(r.json()["results"])
     ```
 
+- ### Gradio App 支持
+  从 PaddleHub 2.3.1 开始支持使用链接 http://127.0.0.1:8866/gradio/chinese_ocr_db_crnn_server 在浏览器中访问 chinese_ocr_db_crnn_server 的 Gradio App。
 
 ## 五、更新历史
 
@@ -178,20 +180,24 @@
 
 * 1.1.0
 
-   使用三阶段模型（文本框检测-角度分类-文字识别）识别图片文字。
+  使用三阶段模型（文本框检测-角度分类-文字识别）识别图片文字。
 
 * 1.1.1
 
-   支持文本中空格识别。
+  支持文本中空格识别。
 
 * 1.1.2
 
-   修复检出字段无法超过30个问题。
+  修复检出字段无法超过30个问题。
 
 * 1.1.3
 
-   移除 fluid api
+  移除 fluid api
+
+* 1.2.0
+
+  添加 Gradio APP
 
   - ```shell
-    $ hub install chinese_ocr_db_crnn_server==1.1.3
+    $ hub install chinese_ocr_db_crnn_server==1.2.0
     ```

modules/image/text_recognition/chinese_ocr_db_crnn_server/character.py

@@ -11,9 +11,9 @@
 # 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 string
 
 import numpy as np
-import string
 
 
 class CharacterOps(object):

modules/image/text_recognition/chinese_ocr_db_crnn_server/module.py

-# -*- coding:utf-8 -*-
 from __future__ import absolute_import
 from __future__ import division
 from __future__ import print_function
@@ -23,23 +22,23 @@ from paddle.inference import create_predictor
 from PIL import Image
 
 import paddlehub as hub
-from paddlehub.common.logger import logger
 from paddlehub.module.module import moduleinfo
 from paddlehub.module.module import runnable
 from paddlehub.module.module import serving
+from paddlehub.utils.log import logger
 
 
 @moduleinfo(
     name="chinese_ocr_db_crnn_server",
-    version="1.1.3",
+    version="1.2.0",
     summary=
     "The module can recognize the chinese texts in an image. Firstly, it will detect the text box positions based on the differentiable_binarization_chn module. Then it recognizes the chinese texts. ",
     author="paddle-dev",
     author_email="paddle-dev@baidu.com",
     type="cv/text_recognition")
-class ChineseOCRDBCRNNServer(hub.Module):
+class ChineseOCRDBCRNNServer:
 
-    def _initialize(self, text_detector_module=None, enable_mkldnn=False):
+    def __init__(self, text_detector_module=None, enable_mkldnn=False):
         """
         initialize with the necessary elements
         """
@@ -57,8 +56,8 @@ class ChineseOCRDBCRNNServer(hub.Module):
         self.font_file = os.path.join(self.directory, 'assets', 'simfang.ttf')
         self.enable_mkldnn = enable_mkldnn
 
-        self.rec_pretrained_model_path = os.path.join(self.directory, 'inference_model', 'character_rec')
-        self.cls_pretrained_model_path = os.path.join(self.directory, 'inference_model', 'angle_cls')
+        self.rec_pretrained_model_path = os.path.join(self.directory, 'inference_model', 'character_rec', 'model')
+        self.cls_pretrained_model_path = os.path.join(self.directory, 'inference_model', 'angle_cls', 'model')
         self.rec_predictor, self.rec_input_tensor, self.rec_output_tensors = self._set_config(
             self.rec_pretrained_model_path)
         self.cls_predictor, self.cls_input_tensor, self.cls_output_tensors = self._set_config(
@@ -68,8 +67,8 @@ class ChineseOCRDBCRNNServer(hub.Module):
         """
         predictor config path
         """
-        model_file_path = os.path.join(pretrained_model_path, 'model')
-        params_file_path = os.path.join(pretrained_model_path, 'params')
+        model_file_path = pretrained_model_path + '.pdmodel'
+        params_file_path = pretrained_model_path + '.pdiparams'
 
         config = Config(model_file_path, params_file_path)
         try:
@@ -111,8 +110,7 @@ class ChineseOCRDBCRNNServer(hub.Module):
         """
         if not self._text_detector_module:
             self._text_detector_module = hub.Module(name='chinese_text_detection_db_server',
-                                                    enable_mkldnn=self.enable_mkldnn,
-                                                    version='1.0.2')
+                                                    enable_mkldnn=self.enable_mkldnn)
         return self._text_detector_module
 
     def read_images(self, paths=[]):
@@ -267,7 +265,7 @@ class ChineseOCRDBCRNNServer(hub.Module):
                         rec_res_final.append({
                             'text': text,
                             'confidence': float(score),
-                            'text_box_position': boxes[index].astype(np.int).tolist()
+                            'text_box_position': boxes[index].astype(np.int64).tolist()
                         })
                 result['data'] = rec_res_final
 
@@ -411,63 +409,46 @@ class ChineseOCRDBCRNNServer(hub.Module):
 
         return rec_res
 
-    def save_inference_model(self, dirname, model_filename=None, params_filename=None, combined=True):
+    def save_inference_model(self, dirname):
         detector_dir = os.path.join(dirname, 'text_detector')
         classifier_dir = os.path.join(dirname, 'angle_classifier')
         recognizer_dir = os.path.join(dirname, 'text_recognizer')
-        self._save_detector_model(detector_dir, model_filename, params_filename, combined)
-        self._save_classifier_model(classifier_dir, model_filename, params_filename, combined)
-        self._save_recognizer_model(recognizer_dir, model_filename, params_filename, combined)
+
+        self._save_detector_model(detector_dir)
+        self._save_classifier_model(classifier_dir)
+        self._save_recognizer_model(recognizer_dir)
         logger.info("The inference model has been saved in the path {}".format(os.path.realpath(dirname)))
 
-    def _save_detector_model(self, dirname, model_filename=None, params_filename=None, combined=True):
-        self.text_detector_module.save_inference_model(dirname, model_filename, params_filename, combined)
+    def _save_detector_model(self, dirname):
+        self.text_detector_module.save_inference_model(dirname)
 
-    def _save_recognizer_model(self, dirname, model_filename=None, params_filename=None, combined=True):
-        if combined:
-            model_filename = "__model__" if not model_filename else model_filename
-            params_filename = "__params__" if not params_filename else params_filename
+    def _save_recognizer_model(self, dirname):
         place = paddle.CPUPlace()
-        exe = paddle.Executor(place)
-
-        model_file_path = os.path.join(self.rec_pretrained_model_path, 'model')
-        params_file_path = os.path.join(self.rec_pretrained_model_path, 'params')
-        program, feeded_var_names, target_vars = paddle.static.load_inference_model(
-            dirname=self.rec_pretrained_model_path,
-            model_filename=model_file_path,
-            params_filename=params_file_path,
-            executor=exe)
-
-        paddle.static.save_inference_model(dirname=dirname,
-                                           main_program=program,
+        exe = paddle.static.Executor(place)
+
+        program, feeded_var_names, target_vars = paddle.static.load_inference_model(self.rec_pretrained_model_path,
+                                                                                    executor=exe)
+        global_block = program.global_block()
+        feed_vars = [global_block.var(item) for item in feeded_var_names]
+        paddle.static.save_inference_model(dirname,
+                                           feed_vars=feed_vars,
+                                           fetch_vars=target_vars,
                                            executor=exe,
-                                           feeded_var_names=feeded_var_names,
-                                           target_vars=target_vars,
-                                           model_filename=model_filename,
-                                           params_filename=params_filename)
-
-    def _save_classifier_model(self, dirname, model_filename=None, params_filename=None, combined=True):
-        if combined:
-            model_filename = "__model__" if not model_filename else model_filename
-            params_filename = "__params__" if not params_filename else params_filename
+                                           program=program)
+
+    def _save_classifier_model(self, dirname):
         place = paddle.CPUPlace()
-        exe = paddle.Executor(place)
-
-        model_file_path = os.path.join(self.cls_pretrained_model_path, 'model')
-        params_file_path = os.path.join(self.cls_pretrained_model_path, 'params')
-        program, feeded_var_names, target_vars = paddle.static.load_inference_model(
-            dirname=self.cls_pretrained_model_path,
-            model_filename=model_file_path,
-            params_filename=params_file_path,
-            executor=exe)
-
-        paddle.static.save_inference_model(dirname=dirname,
-                                           main_program=program,
+        exe = paddle.static.Executor(place)
+
+        program, feeded_var_names, target_vars = paddle.static.load_inference_model(self.cls_pretrained_model_path,
+                                                                                    executor=exe)
+        global_block = program.global_block()
+        feed_vars = [global_block.var(item) for item in feeded_var_names]
+        paddle.static.save_inference_model(dirname,
+                                           feed_vars=feed_vars,
+                                           fetch_vars=target_vars,
                                            executor=exe,
-                                           feeded_var_names=feeded_var_names,
-                                           target_vars=target_vars,
-                                           model_filename=model_filename,
-                                           params_filename=params_filename)
+                                           program=program)
 
     @runnable
     def run_cmd(self, argvs):
@@ -515,3 +496,25 @@ class ChineseOCRDBCRNNServer(hub.Module):
         Add the command input options
         """
         self.arg_input_group.add_argument('--input_path', type=str, default=None, help="diretory to image")
+
+    def create_gradio_app(self):
+        import gradio as gr
+
+        def inference(image, use_gpu=False, box_thresh=0.5, text_thresh=0.5, angle_classification_thresh=0.9):
+            return self.recognize_text(paths=[image],
+                                       use_gpu=use_gpu,
+                                       output_dir=None,
+                                       visualization=False,
+                                       box_thresh=box_thresh,
+                                       text_thresh=text_thresh,
+                                       angle_classification_thresh=angle_classification_thresh)
+
+        return gr.Interface(inference, [
+            gr.Image(type='filepath'),
+            gr.Checkbox(),
+            gr.Slider(0, 1.0, 0.5, step=0.01),
+            gr.Slider(0, 1.0, 0.5, step=0.01),
+            gr.Slider(0, 1.0, 0.5, step=0.01)
+        ], [gr.JSON(label='results')],
+                            title='chinese_ocr_db_crnn_server',
+                            allow_flagging=False)

modules/image/text_recognition/chinese_ocr_db_crnn_server/test.py

+import os
+import shutil
+import unittest
+
+import cv2
+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/KTzZVDjUsXw/download?ixid=MnwxMjA3fDB8MXxzZWFyY2h8MzM3fHx0ZXh0fGVufDB8fHx8MTY2MzUxMTExMQ&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)
+        cls.module = hub.Module(name="chinese_ocr_db_crnn_server")
+
+    @classmethod
+    def tearDownClass(cls) -> None:
+        shutil.rmtree('tests')
+        shutil.rmtree('inference')
+        shutil.rmtree('ocr_result')
+
+    def test_recognize_text1(self):
+        results = self.module.recognize_text(
+            paths=['tests/test.jpg'],
+            use_gpu=False,
+            visualization=False,
+        )
+        self.assertEqual(results[0]['data'], [{
+            'text': 'GIVE.',
+            'confidence': 0.944110095500946,
+            'text_box_position': [[281, 159], [359, 159], [359, 202], [281, 202]]
+        }, {
+            'text': 'THANKS.',
+            'confidence': 0.9850907325744629,
+            'text_box_position': [[258, 199], [382, 199], [382, 240], [258, 240]]
+        }])
+
+    def test_recognize_text2(self):
+        results = self.module.recognize_text(
+            images=[cv2.imread('tests/test.jpg')],
+            use_gpu=False,
+            visualization=False,
+        )
+        self.assertEqual(results[0]['data'], [{
+            'text': 'GIVE.',
+            'confidence': 0.944110095500946,
+            'text_box_position': [[281, 159], [359, 159], [359, 202], [281, 202]]
+        }, {
+            'text': 'THANKS.',
+            'confidence': 0.9850907325744629,
+            'text_box_position': [[258, 199], [382, 199], [382, 240], [258, 240]]
+        }])
+
+    def test_recognize_text3(self):
+        results = self.module.recognize_text(
+            images=[cv2.imread('tests/test.jpg')],
+            use_gpu=True,
+            visualization=False,
+        )
+        self.assertEqual(results[0]['data'], [{
+            'text': 'GIVE.',
+            'confidence': 0.944110095500946,
+            'text_box_position': [[281, 159], [359, 159], [359, 202], [281, 202]]
+        }, {
+            'text': 'THANKS.',
+            'confidence': 0.9850907325744629,
+            'text_box_position': [[258, 199], [382, 199], [382, 240], [258, 240]]
+        }])
+
+    def test_recognize_text4(self):
+        results = self.module.recognize_text(
+            images=[cv2.imread('tests/test.jpg')],
+            use_gpu=False,
+            visualization=True,
+        )
+        self.assertEqual(results[0]['data'], [{
+            'text': 'GIVE.',
+            'confidence': 0.944110095500946,
+            'text_box_position': [[281, 159], [359, 159], [359, 202], [281, 202]]
+        }, {
+            'text': 'THANKS.',
+            'confidence': 0.9850907325744629,
+            'text_box_position': [[258, 199], [382, 199], [382, 240], [258, 240]]
+        }])
+
+    def test_recognize_text5(self):
+        self.assertRaises(AttributeError, self.module.recognize_text, images=['tests/test.jpg'])
+
+    def test_recognize_text6(self):
+        self.assertRaises(AssertionError, self.module.recognize_text, paths=['no.jpg'])
+
+    def test_save_inference_model(self):
+        self.module.save_inference_model('./inference/model')
+
+        self.assertTrue(os.path.exists('./inference/model/angle_classifier.pdmodel'))
+        self.assertTrue(os.path.exists('./inference/model/angle_classifier.pdiparams'))
+
+        self.assertTrue(os.path.exists('./inference/model/text_detector.pdmodel'))
+        self.assertTrue(os.path.exists('./inference/model/text_detector.pdiparams'))
+
+        self.assertTrue(os.path.exists('./inference/model/text_recognizer.pdmodel'))
+        self.assertTrue(os.path.exists('./inference/model/text_recognizer.pdiparams'))
+
+
+if __name__ == "__main__":
+    unittest.main()

modules/image/text_recognition/chinese_ocr_db_crnn_server/utils.py

-# -*- coding:utf-8 -*-
 from __future__ import absolute_import
 from __future__ import division
 from __future__ import print_function
 
+import base64
 import math
+from io import BytesIO
 
-from PIL import Image, ImageDraw, ImageFont
-import base64
 import cv2
 import numpy as np
+from PIL import Image
+from PIL import ImageDraw
+from PIL import ImageFont
 
 
 def draw_ocr(image, boxes, txts, scores, font_file, draw_txt=True, drop_score=0.5):
@@ -174,4 +176,17 @@ def base64_to_cv2(b64str):
     data = base64.b64decode(b64str.encode('utf8'))
     data = np.fromstring(data, np.uint8)
     data = cv2.imdecode(data, cv2.IMREAD_COLOR)
+    if data is None:
+        buf = BytesIO()
+        image_decode = base64.b64decode(b64str.encode('utf8'))
+        image = BytesIO(image_decode)
+        im = Image.open(image)
+        rgb = im.convert('RGB')
+        rgb.save(buf, 'jpeg')
+        buf.seek(0)
+        image_bytes = buf.read()
+        data_base64 = str(base64.b64encode(image_bytes), encoding="utf-8")
+        image_decode = base64.b64decode(data_base64)
+        img_array = np.frombuffer(image_decode, np.uint8)
+        data = cv2.imdecode(img_array, cv2.IMREAD_COLOR)
     return data