Update app.py

app.py

@@ -1,70 +1,73 @@
 import streamlit as st
 import numpy as np
-import tensorflow as tf
-from tensorflow.keras.preprocessing import image
 import cv2
+from tensorflow.keras.preprocessing import image
+import tensorflow as tf
 
 # Load the saved model
 model_path = "fahrnphi_exam_project.keras"
 model = tf.keras.models.load_model(model_path)
 
-
 # Set image dimensions
 img_height, img_width = 150, 150  # Input size for the model
 
-# Define a function for prediction and returning labels and probabilities
+# Define class labels
+class_labels = {0: 'Bell Pepper', 1: 'Carrot', 2: 'Garlic', 3: 'Ginger', 4: 'Jalapeno', 5: 'Onion', 6: 'Potato', 7: 'Sweetpotato', 8: 'Tomato'}
+
+# Define a function to predict labels for multiple regions in an image
 def predict_labels_and_probabilities(image_path):
-    # Load the image using OpenCV
-    img = cv2.imdecode(np.frombuffer(image_path.read(), np.uint8), 1)
-    img_rgb = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
-    
-    # Assuming the input image might contain multiple ingredients, we will process it in patches.
-    # For simplicity, let's divide the image into 4 patches and classify each one
-    h, w, _ = img.shape
-    patches = [
-        img_rgb[0:h//2, 0:w//2],  # Top-left
-        img_rgb[0:h//2, w//2:w],  # Top-right
-        img_rgb[h//2:h, 0:w//2],  # Bottom-left
-        img_rgb[h//2:h, w//2:w],  # Bottom-right
-    ]
-    
+    # Load the image
+    img = image.load_img(image_path)
+    img_array = image.img_to_array(img)
+
+    # Convert to grayscale
+    gray = cv2.cvtColor(img_array, cv2.COLOR_BGR2GRAY)
+
+    # Apply GaussianBlur to reduce noise and improve contour detection
+    blurred = cv2.GaussianBlur(gray, (5, 5), 0)
+
+    # Find edges in the image using Canny edge detection
+    edged = cv2.Canny(blurred, 50, 150)
+
+    # Find contours in the image
+    contours, _ = cv2.findContours(edged.copy(), cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
+
     predictions = []
-    
-    for patch in patches:
-        patch_resized = cv2.resize(patch, (img_height, img_width))
-        patch_array = image.img_to_array(patch_resized)
-        patch_array = np.expand_dims(patch_array, axis=0)
-        patch_array /= 255.  # Scale pixel values
-        
-        preds = model.predict(patch_array)
-        class_idx = np.argmax(preds[0])
-        
-        # Map class indices to class names
-        class_labels = {
-            0: 'Bell Pepper', 1: 'Carrot', 2: 'Garlic', 3: 'Ginger',
-            4: 'Jalapeno', 5: 'Onion', 6: 'Potato', 7: 'Sweetpotato', 8: 'Tomato'
-        }
-        predicted_class = class_labels[class_idx]
-        probability = preds[0][class_idx]
-        
-        predictions.append((predicted_class, probability))
-    
+
+    for contour in contours:
+        # Create a bounding box around each contour
+        x, y, w, h = cv2.boundingRect(contour)
+        if w > 50 and h > 50:  # Filter out small boxes
+            roi = img_array[y:y+h, x:x+w]
+            roi = cv2.resize(roi, (img_height, img_width))
+            roi = np.expand_dims(roi, axis=0)
+            roi = roi / 255.0  # Scale image pixels
+
+            # Predict with the model
+            preds = model.predict(roi)
+            class_idx = np.argmax(preds[0])
+            predicted_class = class_labels[class_idx]
+            probability = preds[0][class_idx]
+
+            predictions.append((predicted_class, probability))
+
     return predictions
 
 # Streamlit App
 st.title("Intelligent Recipe Finder Classification")
 
-uploaded_file = st.file_uploader("Choose an ingredients image...", type="jpg")
+uploaded_file = st.file_uploader("Choose an image with ingredients...", type="jpg")
 
 if uploaded_file is not None:
     # Display the uploaded image
     st.image(uploaded_file, caption='Uploaded Ingredient Image.', use_column_width=True)
 
-    # Perform the prediction and display the results
+    # Predict labels and probabilities for multiple regions
     predictions = predict_labels_and_probabilities(uploaded_file)
-    st.write("Predictions for different patches of the image:")
-    
-    for i, (label, probability) in enumerate(predictions):
-        st.write(f"Patch {i+1}:")
-        st.write("Prediction:", label)
-        st.write("Probability:", probability)
+
+    if predictions:
+        st.write("Predictions:")
+        for i, (label, probability) in enumerate(predictions):
+            st.write(f"{i+1}. **Prediction:** {label} | **Probability:** {probability:.2f}")
+    else:
+        st.write("No significant ingredients detected.")