Upload 4 files

.env

@@ -0,0 +1,4 @@
+project-id=deforestation-detection-459814
+MODEL_URL=https://drive.google.com/uc?id=1fuofy10g5UxPIOqvoo6zVFIgSYjSmCV8
+SERVICE_KEY_ID=render-deforestation-detection@deforestation-detection-459814.iam.gserviceaccount.com
+PRIVATE_KEY={"type": "service_account", "project_id": "deforestation-detection-459814", "private_key_id": "dd01483fd123c11f98dc9bac47e65d0c2382135b", "private_key": "-----BEGIN PRIVATE KEY-----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\nSHJPrEldpqlE5ELa4Kmq8MuQ\n-----END PRIVATE KEY-----\n", "client_email": "render-deforestation-detection@deforestation-detection-459814.iam.gserviceaccount.com", "client_id": "108238598515799130608", "auth_uri": "https://accounts.google.com/o/oauth2/auth", "token_uri": "https://oauth2.googleapis.com/token", "auth_provider_x509_cert_url": "https://www.googleapis.com/oauth2/v1/certs", "client_x509_cert_url": "https://www.googleapis.com/robot/v1/metadata/x509/render-deforestation-detection%40deforestation-detection-459814.iam.gserviceaccount.com", "universe_domain": "googleapis.com"}

app.py

@@ -0,0 +1,27 @@
+from flask import Flask, request, jsonify, render_template
+from backend import run_deforestation_pipeline
+
+app = Flask(__name__)
+
+@app.route('/')
+def index():
+    return render_template("index.html")
+
+@app.route('/predict', methods=['POST'])
+def predict():
+    data = request.get_json()
+
+    result = run_deforestation_pipeline(
+        lat_min=data['lat_min'],
+        lat_max=data['lat_max'],
+        lon_min=data['lon_min'],
+        lon_max=data['lon_max'],
+        start_year=data['start_year'],
+        end_year=data['end_year']
+    )
+
+    return jsonify(result)
+
+if __name__ == '__main__':
+    app.run(host='0.0.0.0', port=5000)
+

backend.py

@@ -0,0 +1,155 @@
+import os
+import ee
+import numpy as np
+import requests
+import io
+import base64
+from rasterio.io import MemoryFile
+import torch
+import segmentation_models_pytorch as smp
+import matplotlib.pyplot as plt
+import gdown
+from dotenv import load_dotenv
+
+load_dotenv()
+
+PRIVATE_KEY = os.getenv("PRIVATE_KEY")
+with open("private-key.json", "w") as f:
+    f.write(PRIVATE_KEY)
+service_account = os.getenv("SERVICE_KEY_ID")
+credentials = ee.ServiceAccountCredentials(service_account, 'private-key.json')
+ee.Initialize(credentials)
+
+MODEL_PATH = "deforestation_unet_full_model.pt"
+MODEL_URL = os.getenv("MODEL_URL")
+
+# Download model only if it doesn't exist
+if not os.path.exists(MODEL_PATH):
+    print("Model not found. Downloading from Google Drive...")
+    gdown.download(MODEL_URL, MODEL_PATH, quiet=False)
+
+# ee.Initialize(project=os.environ["project-id"])
+
+DEVICE = "cuda" if torch.cuda.is_available() else "cpu"
+
+# Load model once
+model = smp.Unet(
+    encoder_name="resnet34",
+    encoder_weights=None,
+    in_channels=4,
+    classes=1,
+    activation=None,
+).to(DEVICE)
+model = torch.load("deforestation_unet_full_model.pt", map_location=DEVICE, weights_only=False)
+model.eval()
+
+def apply_scale_factors(image):
+    optical_bands = image.select('SR_B.').multiply(0.0000275).add(-0.2)
+    thermal_bands = image.select('ST_B.*').multiply(0.00341802).add(149.0)
+    return image.addBands(optical_bands, None, True).addBands(thermal_bands, None, True)
+
+def fetch_rgb_ndvi(region, year, scale=30):
+    start = ee.Date.fromYMD(year, 1, 1)
+    end = ee.Date.fromYMD(year, 12, 31)
+    col = (ee.ImageCollection("LANDSAT/LC08/C02/T1_L2")
+           .filterBounds(region)
+           .filterDate(start, end)
+           .filterMetadata('CLOUD_COVER', 'less_than', 10)
+           .map(apply_scale_factors))
+    image = col.median().clip(region)
+    ndvi = image.normalizedDifference(['SR_B5', 'SR_B4']).rename('NDVI')
+    image = image.addBands(ndvi)
+    return image.select(['SR_B4', 'SR_B3', 'SR_B2']), image.select('NDVI')
+
+def download_geotiff_array(img, region, bands, scale=30):
+    url = img.getThumbURL({
+        'scale': scale,
+        'region': region,
+        'format': 'GeoTIFF',
+        'bands': bands
+    })
+    response = requests.get(url)
+    with MemoryFile(response.content) as memfile:
+        with memfile.open() as src:
+            arr = src.read().astype(np.float32)
+            if arr.max() > 1.5:
+                arr /= 255.0
+    return arr
+
+def predict_from_arrays(rgb_arr, ndvi_arr):
+    rgb_arr = rgb_arr[:3, :, :]
+    ndvi_arr = ndvi_arr[:1, :, :]
+    input_arr = np.concatenate([rgb_arr, ndvi_arr], axis=0)
+    input_tensor = torch.tensor(input_arr).unsqueeze(0).to(DEVICE)
+    with torch.no_grad():
+        pred = torch.sigmoid(model(input_tensor))
+        return (pred > 0.5).float().squeeze().cpu().numpy()
+
+def get_deforestation_color_map(mask_t0, mask_t1):
+    H, W = mask_t0.shape
+    color_map = np.zeros((H, W, 3), dtype=np.uint8)
+
+    retained = (mask_t0 == 1) & (mask_t1 == 1)
+    lost     = (mask_t0 == 1) & (mask_t1 == 0)
+    gained   = (mask_t0 == 0) & (mask_t1 == 1)
+    none     = (mask_t0 == 0) & (mask_t1 == 0)
+
+    color_map[retained] = [0, 255, 0]         # Green
+    color_map[lost]     = [255, 0, 0]         # Red
+    color_map[gained]   = [65, 168, 255]      # Blue (gain)
+    color_map[none]     = [255, 255, 255]     # White (no change)
+
+    return color_map
+
+def run_deforestation_pipeline(lat_min, lat_max, lon_min, lon_max, start_year, end_year):
+    region = ee.Geometry.Rectangle([lon_min, lat_min, lon_max, lat_max])
+
+    rgb_t0_ee, ndvi_t0_ee = fetch_rgb_ndvi(region, start_year)
+    rgb_t0 = download_geotiff_array(rgb_t0_ee, region, ['SR_B4', 'SR_B3', 'SR_B2'])
+    ndvi_t0 = download_geotiff_array(ndvi_t0_ee, region, ['NDVI'])
+
+    rgb_t1_ee, ndvi_t1_ee = fetch_rgb_ndvi(region, end_year)
+    rgb_t1 = download_geotiff_array(rgb_t1_ee, region, ['SR_B4', 'SR_B3', 'SR_B2'])
+    ndvi_t1 = download_geotiff_array(ndvi_t1_ee, region, ['NDVI'])
+
+    mask_t0 = predict_from_arrays(rgb_t0, ndvi_t0)
+    mask_t1 = predict_from_arrays(rgb_t1, ndvi_t1)
+
+    deforested_pixels = ((mask_t0 == 1) & (mask_t1 == 0)).sum()
+    gained_pixels = ((mask_t0 == 0) & (mask_t1 == 1)).sum()
+    total_vegetation_t0 = (mask_t0 == 1).sum()
+
+    percent_loss = (deforested_pixels / total_vegetation_t0) * 100 if total_vegetation_t0 > 0 else 0
+    percent_gain = (gained_pixels / mask_t0.size) * 100  # relative to total area
+
+    color_mask = get_deforestation_color_map(mask_t0, mask_t1)
+
+    # Generate figure in memory
+    fig, axes = plt.subplots(1, 3, figsize=(12, 4))
+
+    axes[0].imshow(mask_t0, cmap="Greens")
+    axes[0].set_title(f"Vegetation in {start_year}")
+    axes[0].axis("off")
+
+    axes[1].imshow(mask_t1, cmap="Greens")
+    axes[1].set_title(f"Vegetation in {end_year}")
+    axes[1].axis("off")
+
+    axes[2].imshow(color_mask)
+    axes[2].set_title(f"Vegetation Change")
+    axes[2].axis("off")
+
+    plt.tight_layout()
+
+    buf = io.BytesIO()
+    plt.savefig(buf, format="png")
+    plt.close(fig)
+    buf.seek(0)
+    img_base64 = base64.b64encode(buf.read()).decode('utf-8')
+
+    return {
+        "percent_deforested": round(percent_loss, 2),
+        "percent_regrowth": round(percent_gain, 2),
+        "image_base64": img_base64
+    }
+

deforestation_unet_full_model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c0552d6fa7db02329b96462065b4c69590b5f5c0daeef32dc5f8ff98d51a83e0
+size 97968507