Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
@@ -16,30 +16,31 @@ from shapely.ops import unary_union
|
|
16 |
from rasterio.features import shapes
|
17 |
import torch
|
18 |
import numpy as np
|
19 |
-
import tempfile
|
20 |
|
21 |
DEVICE = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
22 |
ENCODER = 'se_resnext50_32x4d'
|
23 |
ENCODER_WEIGHTS = 'imagenet'
|
24 |
|
25 |
-
# Define a known temporary directory
|
26 |
-
TEMP_DIR = "/tmp"
|
27 |
|
28 |
-
#model
|
29 |
@st.cache_resource
|
30 |
def load_model():
|
31 |
model = torch.load('deeplabv3 v15.pth', map_location=DEVICE)
|
32 |
model.eval().float()
|
33 |
return model
|
34 |
|
|
|
35 |
best_model = load_model()
|
36 |
|
|
|
37 |
def to_tensor(x, **kwargs):
|
38 |
return x.astype('float32')
|
39 |
|
|
|
40 |
# Preprocessing
|
41 |
preprocessing_fn = smp.encoders.get_preprocessing_fn(ENCODER, ENCODER_WEIGHTS)
|
42 |
|
|
|
43 |
def get_preprocessing(tile_size):
|
44 |
_transform = [
|
45 |
albu.PadIfNeeded(min_height=tile_size, min_width=tile_size, always_apply=True),
|
@@ -49,13 +50,20 @@ def get_preprocessing(tile_size):
|
|
49 |
]
|
50 |
return albu.Compose(_transform)
|
51 |
|
|
|
|
|
|
|
|
|
52 |
def extract_tiles(map_file, model, tile_size=512, overlap=0, batch_size=4, threshold=0.6):
|
|
|
53 |
preprocess = get_preprocessing(tile_size)
|
|
|
54 |
tiles = []
|
55 |
|
56 |
with rasterio.open(map_file) as src:
|
57 |
height = src.height
|
58 |
width = src.width
|
|
|
59 |
effective_tile_size = tile_size - overlap
|
60 |
|
61 |
for y in stqdm(range(0, height, effective_tile_size)):
|
@@ -114,6 +122,7 @@ def extract_tiles(map_file, model, tile_size=512, overlap=0, batch_size=4, thres
|
|
114 |
|
115 |
return tiles
|
116 |
|
|
|
117 |
def create_vector_mask(tiles, output_path):
|
118 |
all_polygons = []
|
119 |
for mask_array, meta in tiles:
|
@@ -139,6 +148,7 @@ def create_vector_mask(tiles, output_path):
|
|
139 |
|
140 |
return gdf, area_m2
|
141 |
|
|
|
142 |
def display_map(shapefile_path, tif_path):
|
143 |
st.title("Map with Shape and TIFF Overlay")
|
144 |
|
@@ -169,10 +179,8 @@ def display_map(shapefile_path, tif_path):
|
|
169 |
# Display the map in Streamlit
|
170 |
m.to_streamlit()
|
171 |
|
|
|
172 |
def main():
|
173 |
-
current_directory = os.getcwd()
|
174 |
-
st.write('current directory:', current_directory)
|
175 |
-
|
176 |
st.title("PV Segmentor")
|
177 |
|
178 |
uploaded_file = st.file_uploader("Choose a TIF file", type="tif")
|
@@ -180,10 +188,14 @@ def main():
|
|
180 |
if uploaded_file is not None:
|
181 |
st.write("File uploaded successfully!")
|
182 |
|
|
|
183 |
resolution = st.radio(
|
184 |
-
|
|
|
|
|
185 |
(512, 1024),
|
186 |
index=0
|
|
|
187 |
)
|
188 |
overlap = st.slider(
|
189 |
'Select the value of overlap',
|
@@ -200,39 +212,37 @@ def main():
|
|
200 |
step=0.01
|
201 |
)
|
202 |
|
203 |
-
st.write('You selected:',
|
204 |
st.write('Selected overlap value:', overlap)
|
205 |
st.write('Selected threshold value:', threshold)
|
206 |
|
|
|
|
|
207 |
if st.button("Process File"):
|
208 |
st.write("Processing...")
|
209 |
|
210 |
-
|
211 |
-
|
212 |
-
temp_filepath = temp_file.name
|
213 |
-
temp_file.write(uploaded_file.getbuffer())
|
214 |
-
|
215 |
-
st.write(f"Temporary file saved at: {temp_filepath}")
|
216 |
|
217 |
best_model.float()
|
218 |
-
tiles = extract_tiles(
|
219 |
|
220 |
st.write("Processing complete!")
|
221 |
|
222 |
-
output_path =
|
223 |
result_gdf, area_m2 = create_vector_mask(tiles, output_path)
|
224 |
|
225 |
st.write("Vector mask created successfully!")
|
226 |
st.write(f"Total area occupied by PV panels: {area_m2:.4f} m^2")
|
227 |
|
228 |
# Offer the shapefile for download
|
229 |
-
shp_files = [f for f in os.listdir(
|
230 |
f.startswith("output_mask") and f.endswith((".shp", ".shx", ".dbf", ".prj"))]
|
231 |
|
232 |
with io.BytesIO() as zip_buffer:
|
233 |
with zipfile.ZipFile(zip_buffer, 'a', zipfile.ZIP_DEFLATED, False) as zip_file:
|
234 |
for file in shp_files:
|
235 |
-
zip_file.write(
|
236 |
|
237 |
zip_buffer.seek(0)
|
238 |
st.download_button(
|
@@ -243,14 +253,13 @@ def main():
|
|
243 |
)
|
244 |
|
245 |
# Display the map with the predicted shapefile
|
246 |
-
display_map(
|
247 |
|
248 |
# Clean up temporary files
|
249 |
-
#os.
|
250 |
-
#st.write(f"Temporary file removed: {temp_filepath}")
|
251 |
#for file in shp_files:
|
252 |
-
|
253 |
-
|
254 |
|
255 |
if __name__ == "__main__":
|
256 |
main()
|
|
|
16 |
from rasterio.features import shapes
|
17 |
import torch
|
18 |
import numpy as np
|
|
|
19 |
|
20 |
DEVICE = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
21 |
ENCODER = 'se_resnext50_32x4d'
|
22 |
ENCODER_WEIGHTS = 'imagenet'
|
23 |
|
|
|
|
|
24 |
|
25 |
+
# Load and prepare the model
|
26 |
@st.cache_resource
|
27 |
def load_model():
|
28 |
model = torch.load('deeplabv3 v15.pth', map_location=DEVICE)
|
29 |
model.eval().float()
|
30 |
return model
|
31 |
|
32 |
+
|
33 |
best_model = load_model()
|
34 |
|
35 |
+
|
36 |
def to_tensor(x, **kwargs):
|
37 |
return x.astype('float32')
|
38 |
|
39 |
+
|
40 |
# Preprocessing
|
41 |
preprocessing_fn = smp.encoders.get_preprocessing_fn(ENCODER, ENCODER_WEIGHTS)
|
42 |
|
43 |
+
|
44 |
def get_preprocessing(tile_size):
|
45 |
_transform = [
|
46 |
albu.PadIfNeeded(min_height=tile_size, min_width=tile_size, always_apply=True),
|
|
|
50 |
]
|
51 |
return albu.Compose(_transform)
|
52 |
|
53 |
+
|
54 |
+
|
55 |
+
|
56 |
+
|
57 |
def extract_tiles(map_file, model, tile_size=512, overlap=0, batch_size=4, threshold=0.6):
|
58 |
+
|
59 |
preprocess = get_preprocessing(tile_size)
|
60 |
+
|
61 |
tiles = []
|
62 |
|
63 |
with rasterio.open(map_file) as src:
|
64 |
height = src.height
|
65 |
width = src.width
|
66 |
+
|
67 |
effective_tile_size = tile_size - overlap
|
68 |
|
69 |
for y in stqdm(range(0, height, effective_tile_size)):
|
|
|
122 |
|
123 |
return tiles
|
124 |
|
125 |
+
|
126 |
def create_vector_mask(tiles, output_path):
|
127 |
all_polygons = []
|
128 |
for mask_array, meta in tiles:
|
|
|
148 |
|
149 |
return gdf, area_m2
|
150 |
|
151 |
+
|
152 |
def display_map(shapefile_path, tif_path):
|
153 |
st.title("Map with Shape and TIFF Overlay")
|
154 |
|
|
|
179 |
# Display the map in Streamlit
|
180 |
m.to_streamlit()
|
181 |
|
182 |
+
|
183 |
def main():
|
|
|
|
|
|
|
184 |
st.title("PV Segmentor")
|
185 |
|
186 |
uploaded_file = st.file_uploader("Choose a TIF file", type="tif")
|
|
|
188 |
if uploaded_file is not None:
|
189 |
st.write("File uploaded successfully!")
|
190 |
|
191 |
+
|
192 |
resolution = st.radio(
|
193 |
+
|
194 |
+
"Selext Processing resolution:",
|
195 |
+
|
196 |
(512, 1024),
|
197 |
index=0
|
198 |
+
|
199 |
)
|
200 |
overlap = st.slider(
|
201 |
'Select the value of overlap',
|
|
|
212 |
step=0.01
|
213 |
)
|
214 |
|
215 |
+
st.write('You selected:',resolution)
|
216 |
st.write('Selected overlap value:', overlap)
|
217 |
st.write('Selected threshold value:', threshold)
|
218 |
|
219 |
+
|
220 |
+
|
221 |
if st.button("Process File"):
|
222 |
st.write("Processing...")
|
223 |
|
224 |
+
with open("temp.tif", "wb") as f:
|
225 |
+
f.write(uploaded_file.getbuffer())
|
|
|
|
|
|
|
|
|
226 |
|
227 |
best_model.float()
|
228 |
+
tiles = extract_tiles("temp.tif", best_model, tile_size=resolution, overlap=overlap, batch_size=4, threshold=threshold)
|
229 |
|
230 |
st.write("Processing complete!")
|
231 |
|
232 |
+
output_path = "output_mask.shp"
|
233 |
result_gdf, area_m2 = create_vector_mask(tiles, output_path)
|
234 |
|
235 |
st.write("Vector mask created successfully!")
|
236 |
st.write(f"Total area occupied by PV panels: {area_m2:.4f} m^2")
|
237 |
|
238 |
# Offer the shapefile for download
|
239 |
+
shp_files = [f for f in os.listdir() if
|
240 |
f.startswith("output_mask") and f.endswith((".shp", ".shx", ".dbf", ".prj"))]
|
241 |
|
242 |
with io.BytesIO() as zip_buffer:
|
243 |
with zipfile.ZipFile(zip_buffer, 'a', zipfile.ZIP_DEFLATED, False) as zip_file:
|
244 |
for file in shp_files:
|
245 |
+
zip_file.write(file)
|
246 |
|
247 |
zip_buffer.seek(0)
|
248 |
st.download_button(
|
|
|
253 |
)
|
254 |
|
255 |
# Display the map with the predicted shapefile
|
256 |
+
display_map("output_mask.shp", "temp.tif")
|
257 |
|
258 |
# Clean up temporary files
|
259 |
+
#os.remove("temp.tif")
|
|
|
260 |
#for file in shp_files:
|
261 |
+
# os.remove(file)
|
262 |
+
|
263 |
|
264 |
if __name__ == "__main__":
|
265 |
main()
|