import gradio as gr import numpy as np import cv2 import tensorflow as tf from PIL import Image # ---------- Load the trained model ---------- smile_model = tf.keras.models.load_model('smile_model.h5') print("✅ Model loaded") # Load Haar Cascade for face detection face_cascade = cv2.CascadeClassifier(cv2.data.haarcascades + 'haarcascade_frontalface_default.xml') # ---------- Detection function (called on each webcam frame) ---------- def detect_smile(image): """ Takes an image (numpy array), detects faces, classifies smile, draws bounding boxes (green=smile, red=no smile), and returns the image. """ # Convert PIL Image or Gradio numpy to OpenCV BGR if isinstance(image, Image.Image): image = np.array(image) img = cv2.cvtColor(image, cv2.COLOR_RGB2BGR) rgb = cv2.cvtColor(img, cv2.COLOR_BGR2RGB) gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY) faces = face_cascade.detectMultiScale(gray, scaleFactor=1.1, minNeighbors=5) for (x, y, w, h) in faces: # Extract face ROI face_roi = rgb[y:y+h, x:x+w] face_resized = cv2.resize(face_roi, (224, 224)) face_array = np.expand_dims(face_resized / 255.0, axis=0) # Predict smile prob = smile_model.predict(face_array, verbose=0)[0][0] label = 'Smile' if prob > 0.5 else 'No Smile' color = (0, 255, 0) if label == 'Smile' else (0, 0, 255) # Draw rectangle and label cv2.rectangle(img, (x, y), (x+w, y+h), color, 2) cv2.putText(img, label, (x, y-10), cv2.FONT_HERSHEY_SIMPLEX, 0.8, color, 2) # Convert back to RGB for Gradio return cv2.cvtColor(img, cv2.COLOR_BGR2RGB) # ---------- Gradio Interface ---------- interface = gr.Interface( fn=detect_smile, inputs=gr.Image(sources=["webcam"], streaming=True, type="numpy", label="Live Webcam"), outputs=gr.Image(type="numpy", label="Detection Result"), live=True, # Real-time updates title="😊 Real-Time Smile Detection", description="Allow camera access and see live smile detection! Green box = Smile, Red box = No Smile." ) if __name__ == "__main__": interface.launch()