import gradio as gr
import cv2
import numpy as np

face_cascade = cv2.CascadeClassifier(
    cv2.data.haarcascades + 'haarcascade_frontalface_default.xml'
)

def mask_image(image):
    """Apply mask to a single image (not streaming)"""
    if image is None:
        return None

    # Convert to BGR for OpenCV
    frame = cv2.cvtColor(image, cv2.COLOR_RGB2BGR)
    gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)

    faces = face_cascade.detectMultiScale(gray, 1.3, 5)

    for (x, y, w, h) in faces:
        # Zorro mask
        cv2.rectangle(frame, (x, y + int(h*0.35)),
                     (x + w, y + int(h*0.65)), (0, 0, 0), -1)
        # Eye holes
        cv2.circle(frame, (x + int(w*0.35), y + int(h*0.5)),
                  int(w*0.1), (255, 255, 255), -1)
        cv2.circle(frame, (x + int(w*0.65), y + int(h*0.5)),
                  int(w*0.1), (255, 255, 255), -1)

    return cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)

# Simple interface - user clicks "Capture" then "Submit"
demo = gr.Interface(
    fn=mask_image,
    inputs=gr.Image(sources=["webcam"], type="numpy"),
    outputs=gr.Image(),
    title="Zorro Mask",
    description="1. Allow camera access\n2. Click 'Capture'\n3. Click 'Submit'\n4. See the mask!",
    examples=None
)

demo.launch()