Update app.py

app.py

@@ -1,93 +1,93 @@
-import cv2
-import gradio as gr
-import numpy as np
-from ultralytics import YOLO
-import tempfile
-import os
-
-# Load YOLOv8 face and plate detection models
-face_model = YOLO("yolov8n-face.pt")
-plate_model = YOLO("icense_plate_detector.pt")
-
-def pixelate(image, boxes, factor=10):
-    for box in boxes:
-        x1, y1, x2, y2 = map(int, box[:4])
-        region = image[y1:y2, x1:x2]
-        h, w = region.shape[:2]
-        if h == 0 or w == 0:
-            continue
-        temp = cv2.resize(region, (factor, factor), interpolation=cv2.INTER_LINEAR)
-        pixelated = cv2.resize(temp, (w, h), interpolation=cv2.INTER_NEAREST)
-        image[y1:y2, x1:x2] = pixelated
-    return image
-
-def detect_and_redact(img, redact_faces, redact_plates):
-    img = cv2.cvtColor(img, cv2.COLOR_RGB2BGR)
-
-    if redact_faces:
-        face_results = face_model.predict(img, conf=0.3, verbose=False)
-        face_boxes = face_results[0].boxes.xyxy.cpu().numpy() if face_results else []
-        img = pixelate(img, face_boxes)
-
-    if redact_plates:
-        plate_results = plate_model.predict(img, conf=0.3, verbose=False)
-        plate_boxes = plate_results[0].boxes.xyxy.cpu().numpy() if plate_results else []
-        img = pixelate(img, plate_boxes)
-
-    return cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
-
-def process_image(input_image, mode):
-    redact_faces = mode in ["Faces Only", "Both"]
-    redact_plates = mode in ["Plates Only", "Both"]
-    return detect_and_redact(input_image, redact_faces, redact_plates)
-
-def process_video(video_file, mode):
-    redact_faces = mode in ["Faces Only", "Both"]
-    redact_plates = mode in ["Plates Only", "Both"]
-
-    cap = cv2.VideoCapture(video_file)
-    fps = cap.get(cv2.CAP_PROP_FPS)
-    width  = int(cap.get(cv2.CAP_PROP_FRAME_WIDTH))
-    height = int(cap.get(cv2.CAP_PROP_FRAME_HEIGHT))
-
-    # Temporary output file
-    temp_video = tempfile.NamedTemporaryFile(suffix=".mp4", delete=False)
-    out = cv2.VideoWriter(temp_video.name, cv2.VideoWriter_fourcc(*'mp4v'), fps, (width, height))
-
-    while cap.isOpened():
-        ret, frame = cap.read()
-        if not ret:
-            break
-        frame_rgb = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)
-        redacted = detect_and_redact(frame_rgb, redact_faces, redact_plates)
-        out.write(cv2.cvtColor(redacted, cv2.COLOR_RGB2BGR))
-
-    cap.release()
-    out.release()
-    return temp_video.name
-
-# Gradio Interface
-with gr.Blocks(title="Visual Privacy Filter") as demo:
-    gr.Markdown("## 🔒 Visual Privacy Filter (Faces & License Plates)")
-    gr.Markdown("Upload an image or video and choose what to redact.")
-
-    mode = gr.Radio(
-        choices=["Faces Only", "Plates Only", "Both"],
-        value="Both",
-        label="Redaction Mode"
-    )
-
-    with gr.Tabs():
-        with gr.Tab("📷 Image"):
-            img_input = gr.Image(type="numpy", label="Upload Image")
-            img_output = gr.Image(type="numpy", label="Redacted Output")
-            img_btn = gr.Button("Redact Image")
-            img_btn.click(fn=process_image, inputs=[img_input, mode], outputs=img_output)
-
-        with gr.Tab("🎥 Video"):
-            vid_input = gr.Video(label="Upload Video")
-            vid_output = gr.Video(label="Redacted Video")
-            vid_btn = gr.Button("Redact Video")
-            vid_btn.click(fn=process_video, inputs=[vid_input, mode], outputs=vid_output)
-
-demo.launch()
+import cv2
+import gradio as gr
+import numpy as np
+from ultralytics import YOLO
+import tempfile
+import os
+
+# Load YOLOv8 face and plate detection models
+face_model = YOLO("yolov8n-face.pt")
+plate_model = YOLO("license_plate_detector.pt")
+
+def pixelate(image, boxes, factor=10):
+    for box in boxes:
+        x1, y1, x2, y2 = map(int, box[:4])
+        region = image[y1:y2, x1:x2]
+        h, w = region.shape[:2]
+        if h == 0 or w == 0:
+            continue
+        temp = cv2.resize(region, (factor, factor), interpolation=cv2.INTER_LINEAR)
+        pixelated = cv2.resize(temp, (w, h), interpolation=cv2.INTER_NEAREST)
+        image[y1:y2, x1:x2] = pixelated
+    return image
+
+def detect_and_redact(img, redact_faces, redact_plates):
+    img = cv2.cvtColor(img, cv2.COLOR_RGB2BGR)
+
+    if redact_faces:
+        face_results = face_model.predict(img, conf=0.3, verbose=False)
+        face_boxes = face_results[0].boxes.xyxy.cpu().numpy() if face_results else []
+        img = pixelate(img, face_boxes)
+
+    if redact_plates:
+        plate_results = plate_model.predict(img, conf=0.3, verbose=False)
+        plate_boxes = plate_results[0].boxes.xyxy.cpu().numpy() if plate_results else []
+        img = pixelate(img, plate_boxes)
+
+    return cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
+
+def process_image(input_image, mode):
+    redact_faces = mode in ["Faces Only", "Both"]
+    redact_plates = mode in ["Plates Only", "Both"]
+    return detect_and_redact(input_image, redact_faces, redact_plates)
+
+def process_video(video_file, mode):
+    redact_faces = mode in ["Faces Only", "Both"]
+    redact_plates = mode in ["Plates Only", "Both"]
+
+    cap = cv2.VideoCapture(video_file)
+    fps = cap.get(cv2.CAP_PROP_FPS)
+    width  = int(cap.get(cv2.CAP_PROP_FRAME_WIDTH))
+    height = int(cap.get(cv2.CAP_PROP_FRAME_HEIGHT))
+
+    # Temporary output file
+    temp_video = tempfile.NamedTemporaryFile(suffix=".mp4", delete=False)
+    out = cv2.VideoWriter(temp_video.name, cv2.VideoWriter_fourcc(*'mp4v'), fps, (width, height))
+
+    while cap.isOpened():
+        ret, frame = cap.read()
+        if not ret:
+            break
+        frame_rgb = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)
+        redacted = detect_and_redact(frame_rgb, redact_faces, redact_plates)
+        out.write(cv2.cvtColor(redacted, cv2.COLOR_RGB2BGR))
+
+    cap.release()
+    out.release()
+    return temp_video.name
+
+# Gradio Interface
+with gr.Blocks(title="Visual Privacy Filter") as demo:
+    gr.Markdown("## 🔒 Visual Privacy Filter (Faces & License Plates)")
+    gr.Markdown("Upload an image or video and choose what to redact.")
+
+    mode = gr.Radio(
+        choices=["Faces Only", "Plates Only", "Both"],
+        value="Both",
+        label="Redaction Mode"
+    )
+
+    with gr.Tabs():
+        with gr.Tab("📷 Image"):
+            img_input = gr.Image(type="numpy", label="Upload Image")
+            img_output = gr.Image(type="numpy", label="Redacted Output")
+            img_btn = gr.Button("Redact Image")
+            img_btn.click(fn=process_image, inputs=[img_input, mode], outputs=img_output)
+
+        with gr.Tab("🎥 Video"):
+            vid_input = gr.Video(label="Upload Video")
+            vid_output = gr.Video(label="Redacted Video")
+            vid_btn = gr.Button("Redact Video")
+            vid_btn.click(fn=process_video, inputs=[vid_input, mode], outputs=vid_output)
+
+demo.launch()