Optimizing the function by loading the model outside predict function

app.py

@@ -6,11 +6,10 @@ from transformers import pipeline
 imageClassifier = pipeline(task="image-classification",
                            model="PranomVignesh/Police-vs-Public")
 
+model = yolov5.load('./best.pt', device="cpu")
 
 def predict(image):
-    model = yolov5.load('./best.pt', device="cpu")
     results = model([image], size=224)
-
     predictions = imageClassifier(image)
     classMappings = {
         'police': "Police / Authorized Personnel",
@@ -23,12 +22,6 @@ def predict(image):
     return results.render()[0], output
 
 
-title = "Detecting Unauthorized Individuals with Firearms"
-
-examples = [
-    []
-]
-
 title = "Detecting Unauthorized Individuals with Firearms"
 description = """
   Try the examples at bottom to get started.
@@ -59,7 +52,6 @@ interface = gr.Interface(
     examples=examples,
     description=description,
     cache_examples=True,
-    live=True,
     theme='huggingface'
 )
 interface.launch(debug=True, share=True, enable_queue=True)