Create app.py

app.py

@@ -0,0 +1,95 @@
+import streamlit as st
+import os
+import numpy as np
+import matplotlib.pyplot as plt
+from tensorflow.keras.models import load_model
+from your_dataset_module import readDataset  # Replace with your actual dataset module
+
+# Configuration
+HEIGHT = WIDTH = 256
+SAR_SHAPE = (HEIGHT, WIDTH, 1)
+OPTIC_SHAPE = (HEIGHT, WIDTH, 3)
+MASK_SHAPE = (HEIGHT, WIDTH, 4)  # One-hot encoded masks with 4 classes
+
+# Class colors: non-mining (green), illegal mining (red), beach (black)
+CLASS_COLORS = [
+    [115/255, 178/255, 115/255],
+    [1, 0, 0],
+    [0, 0, 0]
+]
+
+# Streamlit App Title
+st.title("Satellite Mining Segmentation: SAR + Optic Image Inference")
+
+# Sidebar inputs
+dataset_path = st.text_input("Enter dataset path", "path/to/your/dataset")
+model_path = st.text_input("Enter path to trained model (.h5)", "model.h5")
+num_samples = st.slider("Number of test samples to visualize", 1, 10, 3)
+
+if st.button("Run Inference"):
+    with st.spinner("Loading data and model..."):
+
+        # Prepare paths
+        sar_path = os.path.join(dataset_path, 'sar_images')
+        optic_path = os.path.join(dataset_path, 'optic_images')
+        mask_path = os.path.join(dataset_path, 'masks')
+
+        # Read dataset
+        dataset = readDataset(
+            sarPathes=sar_path,
+            opticPathes=optic_path,
+            masksPathes=mask_path
+        )
+        dataset.readPathes()
+
+        sar_images = dataset.readImages(dataset.sarImages, typeData='s', width=WIDTH, height=HEIGHT)
+        optic_images = dataset.readImages(dataset.opticImages, typeData='o', width=WIDTH, height=HEIGHT)
+        masks = dataset.readImages(dataset.masks, typeData='m', width=WIDTH, height=HEIGHT)
+
+        sar_images = dataset.normalizeImages(sar_images, 's')
+        optic_images = dataset.normalizeImages(optic_images, 'i')
+
+        # Load model
+        model = load_model(model_path)
+
+        # Predict
+        pred_masks = model.predict([optic_images, sar_images], verbose=0)
+        is_multiclass = pred_masks.shape[-1] > 1
+
+        num_samples = min(num_samples, len(sar_images))
+
+        # Plotting
+        fig, axes = plt.subplots(num_samples, 4, figsize=(21, 6 * num_samples))
+
+        for i in range(num_samples):
+            ax = axes[i] if num_samples > 1 else axes
+
+            ax[0].imshow(sar_images[i].squeeze(), cmap='gray')
+            ax[0].set_title(f"SAR Image {i+1}")
+            ax[0].axis('off')
+
+            ax[1].imshow(optic_images[i])
+            ax[1].set_title(f"Optic Image {i+1}")
+            ax[1].axis('off')
+
+            if is_multiclass:
+                gt_color_mask = np.zeros((*masks[i].shape[:2], 3))
+                for j, color in enumerate(CLASS_COLORS):
+                    gt_color_mask += masks[i][:,:,j][:,:,np.newaxis] * np.array(color)
+                ax[2].imshow(gt_color_mask)
+            else:
+                ax[2].imshow(masks[i], cmap='gray')
+            ax[2].set_title(f"Ground Truth Mask {i+1}")
+            ax[2].axis('off')
+
+            if is_multiclass:
+                pred_color_mask = np.zeros((*pred_masks[i].shape[:2], 3))
+                for j, color in enumerate(CLASS_COLORS):
+                    pred_color_mask += pred_masks[i][:,:,j][:,:,np.newaxis] * np.array(color)
+                ax[3].imshow(pred_color_mask)
+            else:
+                ax[3].imshow(pred_masks[i], cmap='gray')
+            ax[3].set_title(f"Predicted Mask {i+1}")
+            ax[3].axis('off')
+
+        st.pyplot(fig)