update requirements and update code to use torch cuda

requirements.txt

@@ -5,4 +5,7 @@ flask-cors==5.0.1
 PyThreadKiller==3.0.6
 vidgear==0.3.3
 selenium==4.32.0
-webdriver_manager==4.0.2
+webdriver_manager==4.0.2
+torch==2.7.0+cu126
+torchaudio==2.7.0+cu126
+torchvision==0.22.0+cu126

src/app.py

@@ -9,12 +9,8 @@ from utils import start_html_stream, get_mjpeg_frames, get_webpage_frames
 import cv2
 from ultralytics import solutions
 import time
+import torch
 
-yolo_detection_model = YOLO('models/yolo11n.pt')
-yolo_segmentation_model = YOLO('models/yolo11n-seg.pt')
-yolo_pose_model = YOLO('models/yolo11n-pose.pt')
-yolo_orientation_model = YOLO('models/yolo11n-obb.pt')
-yolo_classification_model = YOLO('models/yolo11n-cls.pt')
 logging.getLogger('ultralytics').setLevel(logging.WARNING)
 
 PORT = 7860
@@ -96,14 +92,17 @@ def html_stream():
     session_id = f"{stream_url}_{model_type}_{should_track}_{''.join(map(str, classes))}"
 
     model = {
-        'detection': yolo_detection_model,
-        'segmentation': yolo_segmentation_model,
-        'pose': yolo_pose_model,
-        'orientation': yolo_orientation_model,
-        'classification': yolo_classification_model,
-        'regions': 'regions' if regions else yolo_detection_model,
+        'detection': YOLO('models/yolo11n.pt'),
+        'segmentation': YOLO('models/yolo11n-seg.pt'),
+        'pose': YOLO('models/yolo11n-pose.pt'),
+        'orientation': YOLO('models/yolo11n-obb.pt'),
+        'classification': YOLO('models/yolo11n-cls.pt'),
+        'regions': 'regions' if regions else YOLO('models/yolo11n.pt'),
         'preview': None
     }.get(model_type, None)
+
+    device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
+    model.to(device)
     
     if session_id not in sessions:
         sessions[session_id] = {
@@ -164,4 +163,7 @@ def data_stream():
         return f"Session {session_id} not found", 204
     return Response(generate(), mimetype='text/event-stream')
 
+if __name__ == '__main__':
+    print("CUDA Available:", torch.cuda.is_available())
+
 app.run(host="0.0.0.0", port=PORT, debug=True, use_reloader=False)