app.py

# Import libraries
import numpy as np
import tensorflow as tf
from tensorflow.keras.layers import Input, Conv2D, MaxPooling2D, UpSampling2D, concatenate
import matplotlib.pyplot as plt
import rasterio
import os
from huggingface_hub import HfApi, HfFolder
from huggingface_hub import create_repo, upload_file

# Define the U-Net model
def build_unet(input_shape=(128, 128, 1)):
    inputs = Input(input_shape)

    # Encoder
    c1 = Conv2D(64, (3, 3), activation='relu', padding='same')(inputs)
    p1 = MaxPooling2D((2, 2))(c1)

    c2 = Conv2D(128, (3, 3), activation='relu', padding='same')(p1)
    p2 = MaxPooling2D((2, 2))(c2)

    c3 = Conv2D(256, (3, 3), activation='relu', padding='same')(p2)

    # Decoder
    u1 = UpSampling2D((2, 2))(c3)
    m1 = concatenate([u1, c2])
    c4 = Conv2D(128, (3, 3), activation='relu', padding='same')(m1)

    u2 = UpSampling2D((2, 2))(c4)
    m2 = concatenate([u2, c1])
    c5 = Conv2D(64, (3, 3), activation='relu', padding='same')(m2)

    outputs = Conv2D(1, (1, 1), activation='sigmoid')(c5)

    model = tf.keras.Model(inputs, outputs)
    return model

# Prepare the model
input_shape = (128, 128, 1)
unet_model = build_unet(input_shape)
unet_model.compile(optimizer='adam', loss='binary_crossentropy', metrics=['accuracy'])
unet_model.summary()

# Load sample terrain data (replace with your own dataset)
# For demonstration, we'll create synthetic data
def generate_synthetic_data(num_samples=100, img_size=128):
    data = []
    labels = []
    for _ in range(num_samples):
        image = np.random.rand(img_size, img_size, 1)
        label = (image > 0.5).astype(float)
        data.append(image)
        labels.append(label)
    return np.array(data), np.array(labels)

# Generate synthetic data
X_train, y_train = generate_synthetic_data()

# Train the model
history = unet_model.fit(X_train, y_train, epochs=5, batch_size=8, validation_split=0.2)

# Save the model
unet_model.save("unet_flood_model.h5")

# Visualize results
def plot_sample(image, mask, prediction=None):
    plt.figure(figsize=(12, 4))
    plt.subplot(1, 3, 1)
    plt.title("Input")
    plt.imshow(image.squeeze(), cmap='gray')
    plt.subplot(1, 3, 2)
    plt.title("Ground Truth")
    plt.imshow(mask.squeeze(), cmap='gray')
    if prediction is not None:
        plt.subplot(1, 3, 3)
        plt.title("Prediction")
        plt.imshow(prediction.squeeze(), cmap='gray')
    plt.show()

# Test the model
sample_image = X_train[0:1]
sample_mask = y_train[0:1]
prediction = unet_model.predict(sample_image)
plot_sample(sample_image, sample_mask, prediction)



# Define repository name and token
repo_name = "flood-risk-prediction"
hf_token = "colab api"  # Replace with your token

# Create the repository
create_repo(repo_name, token=hf_token, exist_ok=True)

# Upload the model file
upload_file(
    path_or_fileobj="unet_flood_model.h5",  # Path to your model file
    path_in_repo="unet_flood_model.h5",    # Target path in the repository
    repo_id=f"Asrar990/{repo_name}",  # Replace 'your_username' with your Hugging Face username
    token=hf_token                         # Use your token here
)