Update app.py

app.py

@@ -2,7 +2,6 @@ import gradio as gr
 import numpy as np
 import cv2
 from keras.models import load_model
-from collections import deque
 
 # Load the trained model
 model = load_model("bullying_detection_model.keras")
@@ -12,29 +11,13 @@ IMAGE_HEIGHT, IMAGE_WIDTH = 64, 64
 SEQUENCE_LENGTH = 16
 CLASSES_LIST = ["NonBullying", "Bullying"]
 
-# Queue for real-time webcam frames
-frame_queue = deque(maxlen=SEQUENCE_LENGTH)
-
 # Helper to preprocess frames
 def preprocess_frames(frames):
     processed = [cv2.resize(f, (IMAGE_HEIGHT, IMAGE_WIDTH)) / 255.0 for f in frames]
     return np.array(processed)
 
-# Function for webcam real-time prediction
-def predict_webcam(frame):
-    frame_queue.append(frame)
-    if len(frame_queue) < SEQUENCE_LENGTH:
-        return "Collecting frames..."
-    input_frames = preprocess_frames(list(frame_queue))
-    input_frames = np.expand_dims(input_frames, axis=0)
-    preds = model.predict(input_frames, verbose=0)[0]
-    pred_idx = np.argmax(preds)
-    pred_class = CLASSES_LIST[pred_idx]
-    confidence = float(preds[pred_idx])
-    return f"Prediction: {pred_class} (Confidence: {confidence:.2f})"
-
-# Function for uploaded video prediction
-def predict_video(video):
+# Function for uploaded video
+def predict_bullying(video):
     cap = cv2.VideoCapture(video)
     frames = []
     while True:
@@ -43,33 +26,49 @@ def predict_video(video):
             break
         frames.append(frame)
     cap.release()
+
     if len(frames) < SEQUENCE_LENGTH:
         return "Video too short"
+
     # Sample SEQUENCE_LENGTH evenly spaced frames
     idxs = np.linspace(0, len(frames) - 1, SEQUENCE_LENGTH).astype(int)
     sampled_frames = [frames[i] for i in idxs]
     input_frames = preprocess_frames(sampled_frames)
     input_frames = np.expand_dims(input_frames, axis=0)
-    preds = model.predict(input_frames, verbose=0)[0]
+
+    preds = model.predict(input_frames)[0]
     pred_idx = np.argmax(preds)
     pred_class = CLASSES_LIST[pred_idx]
     confidence = float(preds[pred_idx])
     return f"Prediction: {pred_class} (Confidence: {confidence:.2f})"
 
-# Gradio interface with tabs
+# Function for live webcam input (single frame stream)
+def predict_webcam(frame):
+    frame = cv2.cvtColor(frame, cv2.COLOR_RGB2BGR)  # Convert from Gradio (RGB) to OpenCV (BGR)
+    frame_resized = cv2.resize(frame, (IMAGE_HEIGHT, IMAGE_WIDTH)) / 255.0
+    frames = np.repeat(frame_resized[np.newaxis, :, :, :], SEQUENCE_LENGTH, axis=0)  # Fake sequence
+    input_frames = np.expand_dims(frames, axis=0)
+
+    preds = model.predict(input_frames)[0]
+    pred_idx = np.argmax(preds)
+    pred_class = CLASSES_LIST[pred_idx]
+    confidence = float(preds[pred_idx])
+    return {pred_class: confidence}
+
+# Build Gradio UI
 with gr.Blocks() as demo:
-    gr.Markdown("## Bullying Detection: Webcam & Video Upload Demo")
+    gr.Markdown("# 🎥 Real-Time Bullying Detection\nUpload a video or use your webcam.")
+
+    with gr.Tab("Upload Video"):
+        video_input = gr.Video()
+        video_output = gr.Textbox()
+        video_button = gr.Button("Analyze Video")
+        video_button.click(predict_bullying, inputs=video_input, outputs=video_output)
 
     with gr.Tab("Webcam"):
         webcam_output = gr.Label(num_top_classes=1)
-        webcam_input = gr.Camera()
+        webcam_input = gr.Image(source="webcam", streaming=True)
         webcam_input.stream(predict_webcam, outputs=webcam_output)
 
-    with gr.Tab("Video Upload"):
-        video_input = gr.Video()
-        video_output = gr.Label(num_top_classes=1)
-        video_button = gr.Button("Predict")
-        video_button.click(predict_video, inputs=video_input, outputs=video_output)
-
 if __name__ == "__main__":
     demo.launch()