Update app.py

app.py

@@ -2,50 +2,30 @@ import cv2
 import gradio as gr
 from transformers import pipeline
 
-# Load Hugging Face models
-pose_detection = pipeline("object-detection", model="./yolov8-pose")  # Local YOLOv8 pose model
-suspicious_activity_detection = pipeline("text-classification", model="./suspicious_activity_model")  # Local LSTM model
+# Load YOLOv8 model and suspicious activity classification model
+pose_detection = pipeline("object-detection", model="yolov8-pose")  # Correct path if it's inside same folder.
+suspicious_activity_detection = pipeline("text-classification", model="suspicious_activity_model")
 
-# Function to process a single frame and predict suspicious activity
 def process_frame(frame):
-    """
-    Detect persons in the frame and determine if they exhibit suspicious behavior.
-    """
     results = pose_detection(frame)
 
     for person in results:
-        if person['label'] == 'person' and 'box' in person:
+        if person['label'] == 'person':
             x1, y1, x2, y2 = map(int, person['box'].values())
-
-            if 'keypoints' in person:
-                keypoints = person['keypoints']
-                keypoints_input = " ".join(map(str, [kp for point in keypoints for kp in point[:2]]))
-
-                # Predict suspicious activity
-                prediction = suspicious_activity_detection(keypoints_input)[0]['label']
-                color = (0, 0, 255) if prediction == "Suspicious" else (0, 255, 0)
-                label = 'Suspicious' if prediction == "Suspicious" else 'Normal'
-
-                # Annotate frame
-                cv2.rectangle(frame, (x1, y1), (x2, y2), color, 2)
-                cv2.putText(frame, label, (x1, y1 - 10), cv2.FONT_HERSHEY_SIMPLEX, 0.5, color, 2)
-            else:
-                print("No keypoints found for detected person.")
+            cv2.rectangle(frame, (x1, y1), (x2, y2), (0, 255, 0), 2)
+            cv2.putText(frame, 'Detected', (x1, y1 - 10), cv2.FONT_HERSHEY_SIMPLEX, 0.5, (0, 255, 0), 2)
     return frame
 
-# Gradio interface
 def live_detection(frame):
-    frame = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)  # Convert BGR to RGB for Gradio
+    frame = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)
     processed_frame = process_frame(frame)
-    return cv2.cvtColor(processed_frame, cv2.COLOR_RGB2BGR)  # Convert back to BGR for display
+    return processed_frame
 
-# Gradio UI
 interface = gr.Interface(
     fn=live_detection,
     inputs=gr.Image(source="webcam", tool="editor", type="numpy"),
-    outputs=gr.Image(type="numpy", label="Processed Video Stream"),
-    live=True,
-    description="Real-time Suspicious Activity Detection"
+    outputs=gr.Image(type="numpy"),
+    live=True
 )
 
 if __name__ == "__main__":