Update app.py

app.py

@@ -1,9 +1,10 @@
 import streamlit as st
 import numpy as np
 import tensorflow as tf
-from tensorflow.keras.preprocessing.image import load_img, img_to_array
-from PIL import Image
+import rasterio
 import json
+import tempfile
+import os
 
 # Load the model
 model = tf.keras.models.load_model("EuroSAT_model.h5")
@@ -11,38 +12,52 @@ model = tf.keras.models.load_model("EuroSAT_model.h5")
 # Load class labels
 with open("label_map.json", "r") as f:
     class_labels = json.load(f)
-
-# Reverse the class labels to get index to class name mapping
 class_labels = {v: k for k, v in class_labels.items()}
 
-# Function to preprocess the image
-def preprocess_image(image, target_size=(224, 224)):
-    image = image.resize(target_size)
-    image = img_to_array(image)
-    image = np.expand_dims(image, axis=0)
-    image = image / 255.0  # Normalizing
-    return image
+# Function to preprocess a .tif image and compute indices
+def preprocess_tif(file_path, target_size=(224, 224)):
+    with rasterio.open(file_path) as src:
+        B3 = src.read(3).astype(np.float32)  # Green
+        B4 = src.read(4).astype(np.float32)  # Red
+        B5 = src.read(5).astype(np.float32)  # NIR
+        B6 = src.read(6).astype(np.float32)  # SWIR1
+
+        NDVI = (B5 - B4) / (B5 + B4 + 1e-5)
+        NDBI = (B6 - B5) / (B6 + B5 + 1e-5)
+        NDWI = (B3 - B5) / (B3 + B5 + 1e-5)
+
+        index_img = np.stack([NDBI, NDVI, NDWI], axis=-1)
+        index_img = np.nan_to_num(index_img)
+        index_img = tf.image.resize(index_img, target_size).numpy()
+        index_img = np.clip(index_img, -1, 1)
+        index_img = np.expand_dims(index_img, axis=0)  # Add batch dimension
+
+    return index_img
 
 # Streamlit App
-st.title("Satellite Image Classification")
+st.title("Satellite Image Classification with Spectral Indices")
 
-st.write("Upload a satellite image, and the model will predict its class.")
+st.write("Upload a .tif satellite image. The model computes NDVI, NDBI, and NDWI, then predicts the class.")
 
-# Image uploader
-uploaded_file = st.file_uploader("Choose an image...", type=["jpg", "jpeg", "png"])
+# File uploader
+uploaded_file = st.file_uploader("Choose a .tif image...", type=["tif", "tiff"])
 
 if uploaded_file is not None:
-    # Display the uploaded image
-    image = Image.open(uploaded_file)
-    st.image(image, caption="Uploaded Image", use_column_width=True)
-    
-    # Preprocess the image
-    processed_image = preprocess_image(image)
-    
-    # Make prediction
+    # Save uploaded file to a temporary location
+    with tempfile.NamedTemporaryFile(delete=False, suffix=".tif") as tmp:
+        tmp.write(uploaded_file.read())
+        tmp_path = tmp.name
+
+    # Preprocess image
+    processed_image = preprocess_tif(tmp_path)
+
+    # Predict
     predictions = model.predict(processed_image)
     class_idx = np.argmax(predictions, axis=1)[0]
     predicted_class = class_labels[class_idx]
-    
-    # Show the predicted class
-    st.write(f"Predicted Class: {predicted_class}")
+
+    # Show results
+    st.write(f"Predicted Class: **{predicted_class}**")
+
+    # Cleanup temp file
+    os.remove(tmp_path)