oculus_v5.py

import cv2
import mediapipe as mp
import threading
import numpy as np
import pyautogui as gui
import time

gui.FAILSAFE=False

mp_hands=mp.solutions.hands
mp_draw=mp.solutions.drawing_utils

screen_b , screen_h = gui.size()

cap= cv2.VideoCapture(0)
running = True


def listen_stop():
    global running
    while running:
        user_input=input()
        if user_input.lower() == "stop" or user_input.lower() == "exit" or user_input.lower() == "quit":
            runnning = False

threading.Thread(target = listen_stop , daemon=True).start()


def fin_deg(p1,p2,p3):
    
    p1 = np.array([p1.x, p1.y, p1.z])
    p2 = np.array([p2.x, p2.y, p2.z])
    p3 = np.array([p3.x, p3.y, p3.z])

    t1= p1-p2
    t2= p3-p2

    t1_u = t1/np.linalg.norm(t1)
    t2_u = t2/np.linalg.norm(t2)

    dot_t= np.clip(np.dot(t1_u, t2_u), -1.0, 1.0)
    t_rad= np.arccos(dot_t)
    t_deg= np.degrees(t_rad)

    return t_deg



def finger_up(fin,mcp,pip,dip,tip):
    #bhooliyo mat
    #mcp= metacarpophalanges
    #pip= proximal-interphalanges
    #dip= distal-interphalanges
    #tip= finger tip
    
    if tip==4: #thumb ke liye separate logic for accuracy
        tip_angle= fin_deg(fin.landmark[pip],fin.landmark[dip],fin.landmark[tip])
        if tip_angle<=145:
            return 0
        return 1
    
    else:
        pip_angle= fin_deg(fin.landmark[mcp],fin.landmark[pip],fin.landmark[dip])
        #tip_angle= fin_deg(fin.landmark[pip],fin.landmark[dip],fin.landmark[tip])
        #print(f"PIP Angle: {pip_angle:.2f}, TIP Angle: {tip_angle:.2f}")
        if pip_angle>150: #and tip_angle>165:
            return 1
        return 0


def detect_gesture(fingers, ind_x, ind_y, th_x, th_y):

    if fingers == [0,0,0,0,0] :
        drag_start(ind_x+30, ind_y+310)
        return "FIST"
    
    elif fingers == [1,1,1,1,1] :
        drag_stop()
        return "PALM"
    
    elif fingers == [0,1,1,0,0] :
        return "PEACE"
    
    elif fingers == [1,0,0,0,0]:
        return "THUMBS UP"
    
    elif fingers == [0,0,0,0,1]:
        return "EXCUSE ME"
    
    elif fingers == [1,0,0,0,1]:
        return "CALL!"
    
    elif fingers == [1,1,0,0,0]:
        if abs(th_x-ind_x)<=50 and abs(th_y-ind_y)<=60:
            cur_click()
        return "SHOOT"
    
    elif fingers == [0,1,0,0,1] or fingers == [1,1,0,0,1]:
        return "YO"
    
    elif fingers == [0,1,0,0,0]:
        move_cursor(ind_x, ind_y)
        return "ONE"
    
    elif fingers == [0,1,1,1,0]:
        scroll_up()
        return "THREE"

    elif fingers == [0,1,1,1,1]:
        scroll_down()
        return "FOUR"

    else:
        return "UNKNOWN GESTURE"

def move_cursor(pos_x, pos_y):
    gui.moveTo(pos_x, pos_y, duration=0)

def cur_click():
    gui.click()

def scroll_up():
    gui.scroll(100)

def scroll_down():
    gui.scroll(-100)

def drag_start(x,y):
    gui.mouseDown(x,y)

def drag_stop():
    gui.mouseUp()


print("\n----------------------------------------------------------------\nModel Started, Press 'q' to stop!\n----------------------------------------------------------------\n")

with mp_hands.Hands(static_image_mode=False,
                    max_num_hands=1,
                    min_detection_confidence=0.5,
                    min_tracking_confidence=0.5) as hands:
    while running:
        success, img = cap.read()
        if not success:
            print("Camera not detected, jaake theek kar!")
            break

        img = cv2.flip(img,1)
        img_rgb = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
        result = hands.process(img_rgb)
        img_h, img_b, _ = img.shape
        
        if result.multi_hand_landmarks:
            
            for i, hand_landmarks in enumerate(result.multi_hand_landmarks):
                
                world_landmarks = result.multi_hand_world_landmarks[i]

                fingers = [finger_up(world_landmarks,1,2,3,4),
                           finger_up(world_landmarks,5,6,7,8),
                           finger_up(world_landmarks,9,10,11,12),
                           finger_up(world_landmarks,13,14,15,16),
                           finger_up(world_landmarks,17,18,19,20)]

                th_x = (hand_landmarks.landmark[4].x)*screen_b
                th_y = (hand_landmarks.landmark[4].y)*screen_h
                ind_x =(hand_landmarks.landmark[8].x)*screen_b
                ind_y =(hand_landmarks.landmark[8].y)*screen_h
                
                gesture = detect_gesture(fingers, ind_x, ind_y, th_x, th_y)

                cv2.putText(img, gesture, (50,80+i*30), cv2.FONT_HERSHEY_SIMPLEX,
                            1, (8,143,143), 2)

                mp_draw.draw_landmarks(img, hand_landmarks,
                                       mp_hands.HAND_CONNECTIONS,
                                       mp_draw.DrawingSpec(color=(0,0,255),thickness=2, circle_radius=1),
                                       mp_draw.DrawingSpec(color=(0,255,0), thickness=2))
                time.sleep(0.0000001)
        cv2.imshow("Alex's Oculus Model 5", img)
        if cv2.waitKey(1) & 0xFF == ord('q'):
            running = False

cap.release()
cv2.destroyAllWindows()