mouse_controller_v1.py

import cv2
import mediapipe as mp 
import pyautogui
import math
import time # For better cooldowns

# Setup
hand_detector = mp.solutions.hands.Hands(max_num_hands=1, min_detection_confidence=0.7)
drawing_utils = mp.solutions.drawing_utils
screen_width, screen_height = pyautogui.size()
cap = cv2.VideoCapture(0)

# Cooldown variables to avoid spamming shortcuts
last_action_time = 0
action_cooldown = 1.0 # 1 second

while True:
    _, frame = cap.read()
    frame = cv2.flip(frame, 1)
    frame_height, frame_width, _ = frame.shape
    rgb_frame = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)
    output = hand_detector.process(rgb_frame)
    hands = output.multi_hand_landmarks
    
    curr_time = time.time()

    if hands:
        for hand in hands: 
            drawing_utils.draw_landmarks(frame, hand)
            landmarks = hand.landmark
            
            # --- 1. DEFINE FINGER STATES (Up or Down) ---
            # Index(8), Middle(12), Ring(16), Pinky(20), Thumb(4)
            is_thumb_up = landmarks[4].x < landmarks[3].x # For right hand
            is_index_up = landmarks[8].y < landmarks[6].y
            is_middle_up = landmarks[12].y < landmarks[10].y
            is_ring_up = landmarks[16].y < landmarks[14].y
            is_pinky_up = landmarks[20].y < landmarks[18].y

            # --- 2. GET INDEX COORDINATES FOR MOUSE ---
            ix = int(landmarks[8].x * frame_width)
            iy = int(landmarks[8].y * frame_height)
            screen_x = screen_width/frame_width * ix
            screen_y = screen_height/frame_height * iy

            # --- 3. GESTURE LOGIC ---

            # A. REST MODE (All fingers up) -> Do Nothing
            if is_index_up and is_middle_up and is_ring_up and is_pinky_up:
                cv2.putText(frame, "REST MODE", (50, 50), 1, 2, (0, 255, 0), 2)
                
                # Check for Snap Windows (Move full open hand left/right)
                if curr_time - last_action_time > action_cooldown:
                    if landmarks[0].x < 0.3: # Hand on left
                        pyautogui.hotkey('win', 'left')
                        last_action_time = curr_time
                    elif landmarks[0].x > 0.7: # Hand on right
                        pyautogui.hotkey('win', 'right')
                        last_action_time = curr_time

            # B. MOUSE MODE (Only Index UP)
            elif is_index_up and not is_middle_up and not is_ring_up:
                pyautogui.moveTo(screen_x, screen_y)
                
                # Scrolling inside Mouse Mode: 
                # If Index is bent slightly compared to middle joint
                if landmarks[8].y > landmarks[7].y: 
                    pyautogui.scroll(-100) # Scroll Down

            # C. CLICKING (Index and Middle are both UP)
            elif is_index_up and is_middle_up and not is_ring_up:
                # Use distance between Index (8) and Middle (12) to click
                dist = math.hypot(landmarks[8].x - landmarks[12].x, landmarks[8].y - landmarks[12].y)
                if dist < 0.05: # Pinch gesture
                    pyautogui.click()
                    time.sleep(0.1) # Small buffer, much better than 1s

            # D. ALT + TAB (Fist Mode - All fingers down)
            elif not is_index_up and not is_middle_up and not is_ring_up:
                if curr_time - last_action_time > action_cooldown:
                    pyautogui.hotkey('alt', 'tab')
                    last_action_time = curr_time

    cv2.imshow("Live Capture", frame)
    if cv2.waitKey(1) == ord('q'):
        break

cap.release()
cv2.destroyAllWindows()