Public api

app.py

@@ -0,0 +1,40 @@
+import streamlit as st
+import numpy as np
+from PIL import Image
+from merged_model import CompletedModel
+import time
+
+st.set_page_config(layout="wide")
+st.markdown('<center><h1 style="color: blue;">SCAN ID CARD</h1></center>',
+                          unsafe_allow_html=True)
+
+uploaded_file = st.file_uploader("Upload Files", type=['png','jpeg', 'jpg'])
+if uploaded_file is not None:
+  title_container = st.beta_container()
+  col1, col2, col3, col4 = st.beta_columns((0.6, 0.8, 0.8, 1))
+
+  with title_container:       
+    with col1:
+      image = Image.open(uploaded_file) 
+      img_np = np.array(image)
+
+      st.image(image, caption='ID-Card for detection', width=200)
+      # st.write("")
+      # st.write("Running...")
+      model = CompletedModel()
+      start = time.time()
+      img_crop, img_text, result = model.predict(img_np)
+      end = time.time()
+      total_time = end - start
+
+    with col2:
+      st.image(Image.fromarray(img_crop), caption='Cropped image', width=250)
+
+    with col3:
+      st.image(Image.fromarray(img_text), caption='Text recognition', width=250)
+
+    with col4:
+      st.write(result)
+      st.write('Process time: ' + str(round(total_time, 2)) + 'seconds')
+      st.markdown('<h3 style="color: green;">Scan ID-Card done !!!</h3>',
+                          unsafe_allow_html=True)

complete/merged_model.py

@@ -1,20 +1,21 @@
-from complete.detector import Detector
-from complete.recognition import TextRecognition
+from detector import Detector
+from recognition import TextRecognition
 from utils.image_utils import align_image, sort_text
 import cv2
 import time
 import numpy as np
+from copy import deepcopy as dc
 
 
 class CompletedModel(object):
     def __init__(self):
-        self.corner_detection_model = Detector(path_to_model='./config_corner_detection/Corner/model.tflite',
-                                               path_to_labels='./config_corner_detection/label_map.pbtxt',
+        self.corner_detection_model = Detector(path_to_model='[root]/extract-information-from-identity-card/complete/config_corner_detection/model.tflite',
+                                               path_to_labels='[root]/extract-information-from-identity-card/complete/config_corner_detection/label_map.pbtxt',
                                                nms_threshold=0.2, score_threshold=0.3)
-        self.text_detection_model = Detector(path_to_model='./config_text_detection/Text/model.tflite',
-                                             path_to_labels='./config_text_detection/label_map.pbtxt',
+        self.text_detection_model = Detector(path_to_model='[root]/extract-information-from-identity-card/complete/config_text_detection/model.tflite',
+                                             path_to_labels='[root]/extract-information-from-identity-card/complete/config_text_detection/label_map.pbtxt',
                                              nms_threshold=0.2, score_threshold=0.2)
-        self.text_recognition_model = TextRecognition(path_to_checkpoint='./config_text_recognition/transformerocr.pth')
+        self.text_recognition_model = TextRecognition(path_to_checkpoint='[root]/transformerocr.pth') #download link: https://drive.google.com/file/d/1B7b52G0hL6SKpsxP70xfyKhaRPaoWTuJ/view?usp=sharing
 
         # init boxes
         self.id_boxes = None
@@ -39,7 +40,7 @@ def detect_corner(self, image):
 
         # align image
         cropped_img = align_image(image, coordinate_dict)
-        cv2.imwrite('./test.png', cropped_img)
+        # cv2.imwrite('./test.png', cropped_img)
 
         return cropped_img
 
@@ -50,6 +51,8 @@ def detect_text(self, image):
         # sort text boxes according to coordinate
         self.id_boxes, self.name_boxes, self.birth_boxes, self.home_boxes, self.add_boxes = sort_text(detection_boxes, detection_classes)
 
+        return self.text_detection_model.draw(image)
+
     def text_recognition(self, image):
         field_dict = dict()
 
@@ -82,17 +85,19 @@ def crop_and_recog(boxes):
         field_dict['birth'] = result[len(self.name_boxes) + 1]
         field_dict['home'] = ' '.join(result[len(self.name_boxes) + 2: -len(self.home_boxes)])
         field_dict['add'] = ' '.join(result[-len(self.home_boxes):])
-        print(field_dict)
+        return field_dict
 
     def predict(self, image):
         cropped_image = self.detect_corner(image)
-        self.detect_text(cropped_image)
-        self.text_recognition(cropped_image)
-
-
-img = cv2.imread('./test_image/test.jpeg')
-model = CompletedModel()
-start = time.time()
-model.predict(img)
-end = time.time()
-print("Required time: ", end - start)
+        cropped_image_cp = dc(cropped_image)
+        text_image = self.detect_text(cropped_image)
+        return cropped_image_cp, text_image, self.text_recognition(cropped_image)
+
+
+if "__name__" == "__main__":
+    model = CompletedModel()
+    """
+    image = request.args['image_url']
+    """
+    np_image = np.array(image)
+    _, _, text = model.predict(np_image)

complete/requirement.txt

@@ -1,6 +1,14 @@
-vietocr==0.19.0
 torch
 pillow
 tensorflow==2.2.0
 matplotlib
-torchvision
+torchvision
+tf_slim
+vietocr==0.3.5
+uvicorn
+fastapi
+streamlit
+transformers
+chatterbot
+pyngrok
+imageio

complete/serve_model.py

@@ -1,5 +1,6 @@
 from io import BytesIO
 import numpy as np
+from imageio import imread
 from PIL import Image
 from merged_model import CompletedModel
 
@@ -20,13 +21,12 @@ def predict(image: Image.Image):
         model = load_model()
 
     img = np.asarray(image)
-    result = model.predict(img)
+    _, _, result = model.predict(img)
     print(result)
 
     return result
 
 
 def read_image_file(file) -> Image.Image:
-    image = Image.open(BytesIO(file))
-
+    image = imread(file)
     return image

complete/server.py

@@ -2,27 +2,50 @@
 from fastapi import FastAPI, File, UploadFile
 from starlette.responses import RedirectResponse
 from serve_model import *
+from pydantic import BaseModel
 
 app_desc = """<h2>Try this app by uploading any image with `predict/image`</h2>"""
 
 app = FastAPI(title="Tensorflow FastAPI Start Pack", description=app_desc)
 
 
+class Item(BaseModel):
+    file: str
+
+
 @app.get("/", include_in_schema=False)
 async def index():
     return RedirectResponse(url="/docs")
 
 
+@app.post("/predict/image/url")
+async def predict_api(file: Item):
+    try:
+        extension = file.file.split(".")[-1] in ("jpg", "jpeg", "png")
+        if not extension:
+            return "Image must be jpg or png format!"
+
+        image = read_image_file(file.file)
+        prediction = predict(image)
+
+        return prediction
+    except Exception as e:
+        print("Payload Exception" + str(e))
+
+
 @app.post("/predict/image")
 async def predict_api(file: UploadFile = File(...)):
-    extension = file.filename.split(".")[-1] in ("jpg", "jpeg", "png")
-    if not extension:
-        return "Image must be jpg or png format!"
+    try:
+        extension = file.filename.split(".")[-1] in ("jpg", "jpeg", "png")
+        if not extension:
+            return "Image must be jpg or png format!"
 
-    image = read_image_file(await file.read())
-    prediction = predict(image)
+        image = read_image_file(await file.read())
+        prediction = predict(image)
 
-    return prediction
+        return prediction
+    except Exception as e:
+        print("Payload File Exception" + str(e))
 
 
 if __name__ == "__main__":