Spaces:
Running
Running
Update app.py
Browse files
app.py
CHANGED
@@ -6,40 +6,59 @@ import matplotlib.pyplot as plt
|
|
6 |
from io import BytesIO
|
7 |
|
8 |
# Load the trained model
|
9 |
-
model = load_model('model2.h5')
|
10 |
|
11 |
-
# Prediction function for the Gradio app
|
12 |
def predict_and_visualize(img):
|
13 |
-
#
|
14 |
-
|
15 |
-
|
16 |
-
|
17 |
-
|
18 |
-
|
19 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
20 |
|
21 |
-
|
22 |
-
|
23 |
-
|
24 |
-
|
25 |
-
|
26 |
-
|
27 |
-
|
28 |
-
|
29 |
-
|
30 |
-
|
31 |
-
|
32 |
|
33 |
-
|
|
|
|
|
|
|
34 |
|
35 |
-
# Create the Gradio interface
|
36 |
iface = gr.Interface(
|
37 |
fn=predict_and_visualize,
|
38 |
-
inputs=gr.Image(type="pil"
|
39 |
-
outputs=gr.Image(type="pil"
|
40 |
title="MilitarEye: Military Stealth Camouflage Detector",
|
41 |
-
description="
|
|
|
42 |
)
|
43 |
|
44 |
-
# Launch the
|
45 |
iface.launch()
|
|
|
6 |
from io import BytesIO
|
7 |
|
8 |
# Load the trained model
|
9 |
+
model = load_model('model2.h5')
|
10 |
|
|
|
11 |
def predict_and_visualize(img):
|
12 |
+
# Input validation
|
13 |
+
if img is None:
|
14 |
+
raise gr.Error("Please upload an image")
|
15 |
+
|
16 |
+
try:
|
17 |
+
# Convert numpy array to PIL Image if necessary
|
18 |
+
if isinstance(img, np.ndarray):
|
19 |
+
img = Image.fromarray(img)
|
20 |
+
|
21 |
+
# Store the original image size
|
22 |
+
original_size = img.size
|
23 |
+
|
24 |
+
# Convert the input image to the target size expected by the model
|
25 |
+
img_resized = img.resize((224, 224))
|
26 |
+
img_array = np.array(img_resized) / 255.0 # Normalize the image
|
27 |
+
|
28 |
+
# Ensure the image has 3 channels (RGB)
|
29 |
+
if len(img_array.shape) == 2: # Grayscale image
|
30 |
+
img_array = np.stack((img_array,)*3, axis=-1)
|
31 |
+
elif img_array.shape[-1] == 4: # RGBA image
|
32 |
+
img_array = img_array[:, :, :3]
|
33 |
+
|
34 |
+
img_array = np.expand_dims(img_array, axis=0) # Add batch dimension
|
35 |
|
36 |
+
# Make a prediction
|
37 |
+
prediction = model.predict(img_array)
|
38 |
+
|
39 |
+
# Process the prediction
|
40 |
+
predicted_mask = (prediction[0, :, :, 0] * 255).astype(np.uint8)
|
41 |
+
|
42 |
+
# Convert the prediction to a PIL image
|
43 |
+
prediction_image = Image.fromarray(predicted_mask, mode='L')
|
44 |
+
|
45 |
+
# Resize the predicted image back to the original image size
|
46 |
+
prediction_image = prediction_image.resize(original_size, Image.NEAREST)
|
47 |
|
48 |
+
return prediction_image
|
49 |
+
|
50 |
+
except Exception as e:
|
51 |
+
raise gr.Error(f"Error processing image: {str(e)}")
|
52 |
|
53 |
+
# Create the Gradio interface with examples
|
54 |
iface = gr.Interface(
|
55 |
fn=predict_and_visualize,
|
56 |
+
inputs=gr.Image(type="pil", label="Input Image"),
|
57 |
+
outputs=gr.Image(type="pil", label="Predicted Mask"),
|
58 |
title="MilitarEye: Military Stealth Camouflage Detector",
|
59 |
+
description="Upload an image of a military personnel camouflaged in their surroundings. The model will predict the camouflage mask silhouette.",
|
60 |
+
allow_flagging="never"
|
61 |
)
|
62 |
|
63 |
+
# Launch the app
|
64 |
iface.launch()
|