Spaces:
Sleeping
Sleeping
Mehmet Batuhan Duman
commited on
Commit
·
31179bc
1
Parent(s):
1c0fcfa
Changed scan func
Browse files
app.py
CHANGED
@@ -194,7 +194,7 @@ else:
|
|
194 |
# iface = gr.Interface(fn=gradio_process_image, inputs=inputs, outputs=outputs)
|
195 |
# iface.launch()
|
196 |
|
197 |
-
def scanmap(image_path, model,
|
198 |
satellite_image = cv2.imread(image_path)
|
199 |
satellite_image = satellite_image.astype(np.float32) / 255.0
|
200 |
|
@@ -203,6 +203,8 @@ def scanmap(image_path, model, device, threshold=0.5):
|
|
203 |
|
204 |
height, width, channels = satellite_image.shape
|
205 |
|
|
|
|
|
206 |
|
207 |
fig, ax = plt.subplots(1)
|
208 |
ax.imshow(satellite_image)
|
@@ -213,7 +215,6 @@ def scanmap(image_path, model, device, threshold=0.5):
|
|
213 |
for x in range(0, width - window_size[0] + 1, stride):
|
214 |
cropped_window = satellite_image[y:y + window_size[1], x:x + window_size[0]]
|
215 |
cropped_window_torch = torch.tensor(cropped_window.transpose(2, 0, 1), dtype=torch.float32).unsqueeze(0)
|
216 |
-
cropped_window_torch = cropped_window_torch.to(device) # move data to the same device as model
|
217 |
|
218 |
with torch.no_grad():
|
219 |
probabilities = model(cropped_window_torch)
|
@@ -231,6 +232,7 @@ def scanmap(image_path, model, device, threshold=0.5):
|
|
231 |
|
232 |
return output_path
|
233 |
|
|
|
234 |
def process_image(input_image, model, threshold=0.5):
|
235 |
start_time = time.time()
|
236 |
ship_images = scanmap(input_image, model, threshold)
|
|
|
194 |
# iface = gr.Interface(fn=gradio_process_image, inputs=inputs, outputs=outputs)
|
195 |
# iface.launch()
|
196 |
|
197 |
+
def scanmap(image_path, model, threshold=0.5):
|
198 |
satellite_image = cv2.imread(image_path)
|
199 |
satellite_image = satellite_image.astype(np.float32) / 255.0
|
200 |
|
|
|
203 |
|
204 |
height, width, channels = satellite_image.shape
|
205 |
|
206 |
+
# ensure model is in float32 precision
|
207 |
+
model.float()
|
208 |
|
209 |
fig, ax = plt.subplots(1)
|
210 |
ax.imshow(satellite_image)
|
|
|
215 |
for x in range(0, width - window_size[0] + 1, stride):
|
216 |
cropped_window = satellite_image[y:y + window_size[1], x:x + window_size[0]]
|
217 |
cropped_window_torch = torch.tensor(cropped_window.transpose(2, 0, 1), dtype=torch.float32).unsqueeze(0)
|
|
|
218 |
|
219 |
with torch.no_grad():
|
220 |
probabilities = model(cropped_window_torch)
|
|
|
232 |
|
233 |
return output_path
|
234 |
|
235 |
+
|
236 |
def process_image(input_image, model, threshold=0.5):
|
237 |
start_time = time.time()
|
238 |
ship_images = scanmap(input_image, model, threshold)
|