Update app.py

app.py

@@ -1,76 +1,140 @@
-import pandas as pd
-import numpy as np
 import streamlit as st
-import easyocr
-import PIL
-from PIL import Image, ImageDraw
 from streamlit_drawable_canvas import st_canvas
+import cv2
+import numpy as np
+from tensorflow.keras.models import load_model
+from PIL import Image
+import easyocr
+import pandas as pd
+
+# Load the model for Myanmar character recognition
+model = load_model('mm.h5')
+
+# Initialize EasyOCR reader for English
+reader = easyocr.Reader(['en'], gpu=False)
 
-def rectangle(image, result):
-    """Draw rectangles on image based on predicted coordinates."""
-    draw = ImageDraw.Draw(image)
-    for res in result:
-        top_left = tuple(res[0][0])  # top left coordinates as tuple
-        bottom_right = tuple(res[0][2])  # bottom right coordinates as tuple
-        draw.rectangle((top_left, bottom_right), outline="blue", width=2)
-    # Display image on streamlit
-    st.image(image)
+class_lists = [
+   "0",
+   "1",
+   "2",
+   "3",
+   "4",
+   "5",
+   "6",
+   "7",
+   "8",
+   "9",
+    "Ah",
+    "Aha",
+    "au2",
+    "au3",
+    "ay2",
+    "ba_htoat_chite",
+    "ba_kone",
+    "da_htway",
+    "da_out_chite",
+    "da_yay_hmote",
+    "da_yin_kout",
+    "e1",
+    "e2",
+    "eeare",
+    "ga_khi",
+    "ga_nge",
+    "ha",
+    "hsa_lain",
+    "hta_hsin_htu",
+    "hta_wun_beare",
+    "ka_kji",
+    "kha_khway",
+    "la",
+    "la_kji",
+    "ma",
+    "na_kji",
+    "na_ngear",
+    "nga",
+    "nga_kyi",
+    "O",
+    "pa_sout",
+    "pfa_u_htoat",
+    "sah_lone",
+    "ta_thun_lyin_chate",
+    "ta_wun_pu",
+    "tha",
+    "u1",
+    "u2",
+    "un",
+    "wa",
+    "yah_kout",
+    "yah_pet_let",
+    "za_kwear",
+    "za_myin_hsware"
+]
 
-# Main title
-st.title("Get text from image with EasyOCR")
+# Streamlit UI
+st.title('Text and Character Recognizer')
+st.markdown('''
+Select the mode for recognition:
+''')
 
-# Subtitle
-st.markdown("## EasyOCR with Streamlit")
+# Choose mode
+mode = st.radio("Mode", ('English Text Recognition', 'Myanmar Character Recognition'))
 
-# Upload image file or draw
-st.markdown("## Upload an Image or Draw")
-col1, col2 = st.columns(2)
+if mode == 'English Text Recognition':
+    uploaded_file = st.file_uploader("Upload your file here...", key="uploader_english")
+    if uploaded_file is not None:
+        image = Image.open(uploaded_file)
+        st.image(image, caption='Uploaded Image', use_column_width=True)
 
-with col1:
-    file = st.file_uploader("Upload Here", type=['png', 'jpg', 'jpeg'])
+        # EasyOCR to recognize text
+        result = reader.readtext(np.array(image))
+        for detection in result:
+            st.write(f'Detected text: {detection[1]}, Confidence: {detection[2]}')
 
-with col2:
-    # Drawable canvas
-    canvas_result = st_canvas(
-        fill_color="rgba(255, 165, 0, 0.3)", 
-        stroke_width=3,
-        stroke_color="#ffffff",
-        background_color="#000000",
-        background_image=None if file else st.session_state.get("background", None),
-        update_streamlit=True,
-        width=400,
-        height=400,
-        drawing_mode="freedraw",
-        key="canvas",
-    )
+elif mode == 'Myanmar Character Recognition':
+    col1, col2 = st.columns(2)
 
-# Process uploaded image or drawing
-if file is not None:
-    image = Image.open(file)  # Read image with PIL library
-elif canvas_result.image_data is not None:
-    image = Image.fromarray(canvas_result.image_data.astype('uint8'), 'RGBA').convert('RGB')
-else:
-    st.write("Please upload an image or use the canvas to draw.")
-    image = None
+    with col1:
+        uploaded_file = st.file_uploader("Upload your file here...", key="uploader_myanmar")
 
-if image is not None:
-    st.image(image)  # Display
+    with col2:
+        # Initialize canvas
+        canvas_result = st_canvas(
+            fill_color="rgba(255, 165, 0, 0.3)",
+            stroke_width=3,
+            stroke_color="#ffffff",
+            background_color="#000000",
+            update_streamlit=True,
+            width=200,
+            height=200,
+            drawing_mode="freedraw",
+            key="canvas",
+        )
 
-    # Only detect the English and Turkish part of the image as text
-    reader = easyocr.Reader(['en','ja'], gpu=False) 
-    result = reader.readtext(np.array(image))  # Turn image to numpy array
+    # Process the image for prediction
+    image_data = None
+    if uploaded_file is not None:
+        image_data = Image.open(uploaded_file).convert('RGB')
+    elif canvas_result.image_data is not None:
+        image_data = Image.fromarray(np.uint8(canvas_result.image_data)).convert('RGB')
 
-    # Print all predicted text:
-    for idx, res in enumerate(result):
-        pred_text = res[1]
-        st.write(pred_text)
-    
-    # Collect the results in the dictionary:
-    textdic_easyocr = {res[1]: {'pred_confidence': res[2]} for res in result}
+    if image_data is not None:
+        # Convert PIL image to OpenCV format
+        image_cv = np.array(image_data)
+        image_cv = cv2.cvtColor(image_cv, cv2.COLOR_RGB2BGR)
+        resized_image = cv2.resize(image_cv, (200, 200))
+        # Prepare image for model input
+        model_input = resized_image[np.newaxis, :, :, :3]
 
-    # Create a data frame which shows the predicted text and prediction confidence
-    df = pd.DataFrame.from_dict(textdic_easyocr).T
-    st.table(df)
+        st.write('Model Input')
+        st.image(model_input, width=200)  # Display the input image to model
 
-    # Get boxes on the image 
-    rectangle(image, result)
+        if st.button('Predict Myanmar Character'):
+            # Predict the class
+            val = model.predict(model_input)
+            predicted_class_index = np.argmax(val)
+            mm_text = class_lists[predicted_class_index]
+            st.write(f'Result: {mm_text}, Index: {predicted_class_index}')
+            st.bar_chart(val[0])
+    else:
+        if mode == 'Myanmar Character Recognition':
+            st.write("Please upload an image or draw in the canvas above.")