Update app.py
Browse files
app.py
CHANGED
@@ -12,25 +12,26 @@ import os
|
|
12 |
import csv
|
13 |
import datetime
|
14 |
import zipfile
|
|
|
15 |
|
16 |
-
# === ADMIN
|
17 |
-
ADMIN_KEY = "
|
18 |
|
19 |
-
#
|
20 |
device = torch.device("cpu")
|
21 |
|
22 |
-
#
|
23 |
model = models.resnet50(weights=None)
|
24 |
model.fc = torch.nn.Linear(model.fc.in_features, 2)
|
25 |
model.load_state_dict(torch.load("resnet50_dr_classifier.pth", map_location=device))
|
26 |
model.to(device)
|
27 |
model.eval()
|
28 |
|
29 |
-
#
|
30 |
target_layer = model.layer4[-1]
|
31 |
cam = GradCAM(model=model, target_layers=[target_layer])
|
32 |
|
33 |
-
#
|
34 |
transform = transforms.Compose([
|
35 |
transforms.Resize((224, 224)),
|
36 |
transforms.ToTensor(),
|
@@ -38,7 +39,7 @@ transform = transforms.Compose([
|
|
38 |
[0.229, 0.224, 0.225])
|
39 |
])
|
40 |
|
41 |
-
#
|
42 |
image_folder = "collected_images"
|
43 |
os.makedirs(image_folder, exist_ok=True)
|
44 |
|
@@ -48,7 +49,7 @@ if not os.path.exists(csv_log_path):
|
|
48 |
writer = csv.writer(f)
|
49 |
writer.writerow(["timestamp", "image_filename", "prediction", "confidence"])
|
50 |
|
51 |
-
# ===
|
52 |
def predict_retinopathy(image):
|
53 |
timestamp = datetime.datetime.now().strftime("%Y%m%d_%H%M%S")
|
54 |
img = image.convert("RGB").resize((224, 224))
|
@@ -81,10 +82,7 @@ def predict_retinopathy(image):
|
|
81 |
|
82 |
return cam_pil, f"{label} (Confidence: {confidence:.2f})"
|
83 |
|
84 |
-
# === ADMIN
|
85 |
-
def unlock_downloads(key):
|
86 |
-
return gr.update(visible=True) if key == ADMIN_KEY else gr.update(visible=False)
|
87 |
-
|
88 |
def download_csv():
|
89 |
return csv_log_path
|
90 |
|
@@ -97,9 +95,14 @@ def download_dataset_zip():
|
|
97 |
zipf.write(fpath, arcname=os.path.join("images", fname))
|
98 |
return zip_filename
|
99 |
|
100 |
-
# ===
|
|
|
|
|
|
|
|
|
|
|
101 |
with gr.Blocks() as demo:
|
102 |
-
gr.Markdown("## 🧠 Diabetic Retinopathy Detection with Grad-CAM
|
103 |
|
104 |
with gr.Row():
|
105 |
image_input = gr.Image(type="pil", label="Upload Retinal Image")
|
@@ -114,23 +117,20 @@ with gr.Blocks() as demo:
|
|
114 |
outputs=[cam_output, prediction_output]
|
115 |
)
|
116 |
|
117 |
-
gr.
|
118 |
-
|
119 |
-
with gr.Row():
|
120 |
-
admin_input = gr.Text(label="Enter Admin Key", type="password", placeholder="Only Rodiyah knows this 🔐")
|
121 |
-
unlock_btn = gr.Button("Unlock Downloads")
|
122 |
-
|
123 |
-
with gr.Column(visible=False) as download_section:
|
124 |
with gr.Row():
|
125 |
download_csv_btn = gr.Button("📄 Download CSV Log")
|
126 |
download_zip_btn = gr.Button("📦 Download Full Dataset")
|
127 |
csv_file = gr.File()
|
128 |
zip_file = gr.File()
|
129 |
|
130 |
-
|
131 |
-
|
132 |
-
|
133 |
-
|
|
|
|
|
134 |
)
|
135 |
|
136 |
download_csv_btn.click(
|
|
|
12 |
import csv
|
13 |
import datetime
|
14 |
import zipfile
|
15 |
+
from gradio.routes import Request
|
16 |
|
17 |
+
# === SECRET ADMIN KEY ===
|
18 |
+
ADMIN_KEY = "Diabetes_Detection"
|
19 |
|
20 |
+
# === DEVICE SETUP ===
|
21 |
device = torch.device("cpu")
|
22 |
|
23 |
+
# === MODEL LOADING ===
|
24 |
model = models.resnet50(weights=None)
|
25 |
model.fc = torch.nn.Linear(model.fc.in_features, 2)
|
26 |
model.load_state_dict(torch.load("resnet50_dr_classifier.pth", map_location=device))
|
27 |
model.to(device)
|
28 |
model.eval()
|
29 |
|
30 |
+
# === GRAD-CAM ===
|
31 |
target_layer = model.layer4[-1]
|
32 |
cam = GradCAM(model=model, target_layers=[target_layer])
|
33 |
|
34 |
+
# === IMAGE TRANSFORM ===
|
35 |
transform = transforms.Compose([
|
36 |
transforms.Resize((224, 224)),
|
37 |
transforms.ToTensor(),
|
|
|
39 |
[0.229, 0.224, 0.225])
|
40 |
])
|
41 |
|
42 |
+
# === STORAGE ===
|
43 |
image_folder = "collected_images"
|
44 |
os.makedirs(image_folder, exist_ok=True)
|
45 |
|
|
|
49 |
writer = csv.writer(f)
|
50 |
writer.writerow(["timestamp", "image_filename", "prediction", "confidence"])
|
51 |
|
52 |
+
# === PREDICT FUNCTION ===
|
53 |
def predict_retinopathy(image):
|
54 |
timestamp = datetime.datetime.now().strftime("%Y%m%d_%H%M%S")
|
55 |
img = image.convert("RGB").resize((224, 224))
|
|
|
82 |
|
83 |
return cam_pil, f"{label} (Confidence: {confidence:.2f})"
|
84 |
|
85 |
+
# === ADMIN FILES ===
|
|
|
|
|
|
|
86 |
def download_csv():
|
87 |
return csv_log_path
|
88 |
|
|
|
95 |
zipf.write(fpath, arcname=os.path.join("images", fname))
|
96 |
return zip_filename
|
97 |
|
98 |
+
# === VISIBILITY CHECK ===
|
99 |
+
def is_admin(request: Request):
|
100 |
+
query_params = dict(request.query_params)
|
101 |
+
return query_params.get("admin", "") == ADMIN_KEY
|
102 |
+
|
103 |
+
# === GRADIO APP ===
|
104 |
with gr.Blocks() as demo:
|
105 |
+
gr.Markdown("## 🧠 Diabetic Retinopathy Detection with Grad-CAM + Private Logging")
|
106 |
|
107 |
with gr.Row():
|
108 |
image_input = gr.Image(type="pil", label="Upload Retinal Image")
|
|
|
117 |
outputs=[cam_output, prediction_output]
|
118 |
)
|
119 |
|
120 |
+
with gr.Column(visible=False) as admin_section:
|
121 |
+
gr.Markdown("### 🔐 Admin Downloads")
|
|
|
|
|
|
|
|
|
|
|
122 |
with gr.Row():
|
123 |
download_csv_btn = gr.Button("📄 Download CSV Log")
|
124 |
download_zip_btn = gr.Button("📦 Download Full Dataset")
|
125 |
csv_file = gr.File()
|
126 |
zip_file = gr.File()
|
127 |
|
128 |
+
# Admin visibility only for Rodiyah with key
|
129 |
+
demo.load(
|
130 |
+
lambda req: gr.update(visible=True) if is_admin(req) else gr.update(visible=False),
|
131 |
+
inputs=None,
|
132 |
+
outputs=admin_section,
|
133 |
+
queue=False
|
134 |
)
|
135 |
|
136 |
download_csv_btn.click(
|