import gradio as gr
import cv2
import mediapipe as mp
import numpy as np
import os

mp_hands = mp.solutions.hands
hands = mp_hands.Hands(max_num_hands=1)
slides = sorted([f"slides/{img}" for img in os.listdir("slides")])
index = 0

def detect(frame):
    global index
    h, w, c = frame.shape
    frame_rgb = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)
    result = hands.process(frame_rgb)

    if result.multi_hand_landmarks:
        lm = result.multi_hand_landmarks[0]
        x = lm.landmark[mp_hands.HandLandmark.WRIST].x * w
        if x < w * 0.3:
            index = max(0, index - 1)
        elif x > w * 0.7:
            index = min(len(slides)-1, index + 1)

    slide = cv2.imread(slides[index])
    slide = cv2.cvtColor(slide, cv2.COLOR_BGR2RGB)
    return slide

demo = gr.Interface(
    fn=detect,
    live=True,
    inputs=gr.Image(source="webcam", streaming=True),
    outputs="image"
)

demo.launch()