PaddlePaddle · rainyfly · Dec 29, 2022 · Dec 8, 2022
diff --git a/modules/image/text_recognition/ch_pp-ocrv3/README.md b/modules/image/text_recognition/ch_pp-ocrv3/README.md
@@ -163,6 +163,9 @@
     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/ch_pp-ocrv3 在浏览器中访问 ch_pp-ocrv3 的 Gradio App。
+
 ## 五、更新历史
 
 * 1.0.0
@@ -173,6 +176,10 @@
 
   移除 Fluid API
 
+* 1.2.0
+
+  添加 Gradio APP 支持
+
   - ```shell
-    $ hub install ch_pp-ocrv3==1.1.0
+    $ hub install ch_pp-ocrv3==1.2.0
     ```
diff --git a/modules/image/text_recognition/ch_pp-ocrv3/module.py b/modules/image/text_recognition/ch_pp-ocrv3/module.py
@@ -31,21 +31,22 @@
 from .utils import draw_ocr
 from .utils import get_image_ext
 from .utils import sorted_boxes
-from paddlehub.utils.utils import logger
 from paddlehub.module.module import moduleinfo
 from paddlehub.module.module import runnable
 from paddlehub.module.module import serving
+from paddlehub.utils.utils import logger
 
 
 @moduleinfo(
     name="ch_pp-ocrv3",
-    version="1.1.0",
+    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 classifies the text angle and recognizes the chinese texts. ",
     author="paddle-dev",
     author_email="[email protected]",
     type="cv/text_recognition")
 class ChPPOCRv3:
+
     def __init__(self, text_detector_module=None, enable_mkldnn=False):
         """
         initialize with the necessary elements
@@ -477,3 +478,25 @@ def add_module_input_arg(self):
         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='ch_pp-ocrv3',
+                            allow_flagging=False)
diff --git a/modules/image/text_recognition/ch_pp-ocrv3/requirements.txt b/modules/image/text_recognition/ch_pp-ocrv3/requirements.txt
diff --git a/modules/image/text_recognition/ch_pp-ocrv3/test.py b/modules/image/text_recognition/ch_pp-ocrv3/test.py
@@ -4,13 +4,14 @@
 
 import cv2
 import requests
-import paddlehub as hub
 
+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'
@@ -34,13 +35,13 @@ def test_recognize_text1(self):
             use_gpu=False,
             visualization=False,
         )
-        self.assertEqual(results[0]['data'], [
-        {
-            'text': 'GIVE.', 'confidence': 0.9509806632995605, 
+        self.assertEqual(results[0]['data'], [{
+            'text': 'GIVE.',
+            'confidence': 0.9509768486022949,
             'text_box_position': [[283, 162], [352, 162], [352, 202], [283, 202]]
-        }, 
-        {
-            'text': 'THANKS', 'confidence': 0.9943129420280457, 
+        }, {
+            'text': 'THANKS',
+            'confidence': 0.9943074584007263,
             'text_box_position': [[261, 202], [376, 202], [376, 239], [261, 239]]
         }])
 
@@ -50,13 +51,13 @@ def test_recognize_text2(self):
             use_gpu=False,
             visualization=False,
         )
-        self.assertEqual(results[0]['data'], [
-        {
-            'text': 'GIVE.', 'confidence': 0.9509806632995605, 
+        self.assertEqual(results[0]['data'], [{
+            'text': 'GIVE.',
+            'confidence': 0.9509768486022949,
             'text_box_position': [[283, 162], [352, 162], [352, 202], [283, 202]]
-        }, 
-        {
-            'text': 'THANKS', 'confidence': 0.9943129420280457, 
+        }, {
+            'text': 'THANKS',
+            'confidence': 0.9943074584007263,
             'text_box_position': [[261, 202], [376, 202], [376, 239], [261, 239]]
         }])
 
@@ -66,13 +67,13 @@ def test_recognize_text3(self):
             use_gpu=True,
             visualization=False,
         )
-        self.assertEqual(results[0]['data'], [
-        {
-            'text': 'GIVE.', 'confidence': 0.9509806632995605, 
+        self.assertEqual(results[0]['data'], [{
+            'text': 'GIVE.',
+            'confidence': 0.9509768486022949,
             'text_box_position': [[283, 162], [352, 162], [352, 202], [283, 202]]
-        }, 
-        {
-            'text': 'THANKS', 'confidence': 0.9943129420280457, 
+        }, {
+            'text': 'THANKS',
+            'confidence': 0.9943074584007263,
             'text_box_position': [[261, 202], [376, 202], [376, 239], [261, 239]]
         }])
 
@@ -82,29 +83,21 @@ def test_recognize_text4(self):
             use_gpu=False,
             visualization=True,
         )
-        self.assertEqual(results[0]['data'], [
-        {
-            'text': 'GIVE.', 'confidence': 0.9509806632995605, 
+        self.assertEqual(results[0]['data'], [{
+            'text': 'GIVE.',
+            'confidence': 0.9509768486022949,
             'text_box_position': [[283, 162], [352, 162], [352, 202], [283, 202]]
-        }, 
-        {
-            'text': 'THANKS', 'confidence': 0.9943129420280457, 
+        }, {
+            'text': 'THANKS',
+            'confidence': 0.9943074584007263,
             'text_box_position': [[261, 202], [376, 202], [376, 239], [261, 239]]
         }])
 
     def test_recognize_text5(self):
-        self.assertRaises(
-            AttributeError,
-            self.module.recognize_text,
-            images=['tests/test.jpg']
-        )
+        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']
-        )
+        self.assertRaises(AssertionError, self.module.recognize_text, paths=['no.jpg'])
 
     def test_save_inference_model(self):
         self.module.save_inference_model('./inference/model')

diff --git a/modules/image/text_recognition/ch_pp-ocrv3/utils.py b/modules/image/text_recognition/ch_pp-ocrv3/utils.py
@@ -18,6 +18,7 @@
 
 import base64
 import math
+from io import BytesIO
 
 import cv2
 import numpy as np
@@ -189,4 +190,17 @@ 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