final app before bed

app.py

@@ -1,10 +1,7 @@
 import os
-import cv2
-import numpy as np
-import gradio as gr
-import tensorflow as tf
-import mediapipe as mp
 
+import mediapipe as mp
+import tensorflow as tf
 
 N_ROWS = 543
 N_DIMS = 3
@@ -185,21 +182,34 @@ def extract_keypoints(results):
 # Make prediction
 def make_prediction(processed_landmarks):
     inputs = np.array(processed_landmarks, dtype=np.float32)
+
+    # Set the input tensor for the TFLite model
     interpreter.set_tensor(input_details[0]['index'], inputs)
+
+    # Invoke the TFLite interpreter to perform inference
     interpreter.invoke()
+
+    # Get the output tensor of the TFLite model
     output_data = interpreter.get_tensor(output_details[0]['index'])
+
+    # Find the index of the predicted class
     index = np.argmax(output_data)
-    return inv_index_to_class[index]
+
+    # Map the index to the corresponding class label using the index_to_class dictionary
+    prediction = inv_index_to_class[index]
+
+    return prediction
 
 
 # ...
 
 with mp_holistic.Holistic(min_detection_confidence=0.5, min_tracking_confidence=0.5) as holistic:
-    
+    import cv2
+    import numpy as np
+    import gradio as gr
+    import tensorflow as tf
 
 
-    # ... (Previous code remains the same)
-
     # Modify the predict_with_webcam function to take an image as input and return the prediction string
     def predict_with_webcam(frame):
         if frame is None:
@@ -227,12 +237,6 @@ with mp_holistic.Holistic(min_detection_confidence=0.5, min_tracking_confidence=
             return "Could not detect landmarks. Make sure your webcam is working properly."
 
 
-    # Set mediapipe model
-    cap = cv2.VideoCapture(0)
-
-    cap.release()
-    cv2.destroyAllWindows()
-
     # Define the Gradio interface
     iface = gr.Interface(
         fn=predict_with_webcam,  # The function to use for prediction