Update app.py

app.py

@@ -4,30 +4,16 @@ import keras.backend as K
 import cv2
 import numpy as np
 
-# # Define the custom loss function
-# def ctc_lambda_func(args):
-#     y_pred, labels, input_length, label_length = args
-#     return K.ctc_batch_cost(labels, y_pred, input_length, label_length)
-
-# # Load the model, providing the custom loss function in the custom_objects dictionary
-# loaded_model = load_model('Text_recognizer_Using_CRNN_model.keras', custom_objects={'ctc': ctc_lambda_func},safe_mode=False)
-char_list = "!\"#&'()*+,-./0123456789:;?ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz"
-
-model = load_model('Text_recognizer_Using_CRNN.h5')
-
+try:
+  model = load_model('Text_recognizer_Using_CRNN.h5')
+except Exception as e:
+  print("Not able to load the trained model.")
 
 def process_image(img):
     """
     Converts image to shape (32, 128, 1) & normalize
     """
     w, h = img.shape
-
-#     _, img = cv2.threshold(img,
-#                            128,
-#                            255,
-#                            cv2.THRESH_BINARY | cv2.THRESH_OTSU)
-
-    # Aspect Ratio Calculation
     new_w = 32
     new_h = int(h * (new_w / w))
     img = cv2.resize(img, (new_h, new_w))
@@ -59,23 +45,28 @@ def process_image(img):
 
     return img
 
-filepath="test_img.png"
-img = cv2.imread(filepath, cv2.IMREAD_GRAYSCALE)
-img = process_image(img)
-prediction =model.predict(np.asarray([img]))
-decoded = K.ctc_decode(prediction,
-                       input_length=np.ones(prediction.shape[0]) * prediction.shape[1],
-                       greedy=True)[0][0]
-out = K.get_value(decoded)
-# see the results
-temp_str=""
-for i, x in enumerate(out):
-    for p in x:
-        if int(p) != -1:
-            temp_str+=char_list[int(p)]
+def predict_image_text(input_img):
+  try:
+    img = cv2.cvtColor(input_img, cv2.COLOR_RGB2GRAY)
+    img = process_image(img)
+    prediction =model.predict(np.asarray([img]))
+    decoded = K.ctc_decode(prediction,
+                          input_length=np.ones(prediction.shape[0]) * prediction.shape[1],
+                          greedy=True)[0][0]
+    out = K.get_value(decoded)
+    char_list = "!\"#&'()*+,-./0123456789:;?ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz"
+
+    # see the results
+    temp_text=""
+    for i, x in enumerate(out):
+        for p in x:
+            if int(p) != -1:
+                temp_text+=char_list[int(p)]
+    return temp_text
+  except:
+    return "Some error occured please try again with proper image."
 
-def greet(name):
-    return "Hello " + temp_str + "!!"
+import gradio as gr
 
-demo = gr.Interface(fn=greet, inputs="text", outputs="text")
-demo.launch()
+demo = gr.Interface(predict_image_text, gr.Image(), "text",title="Image to Text Conversion",description="Upload an image and get the extracted text.",theme="compact",)
+demo.launch(debug=True)