main.py

import cv2
import numpy as np
import mediapipe as mp
from mediapipe.tasks import python
from mediapipe.tasks.python import vision

# STEP 1: Configurar o detector no modo VIDEO
base_options = python.BaseOptions(model_asset_path="hand_landmarker.task")
options = vision.HandLandmarkerOptions(
    base_options=base_options,
    running_mode=vision.RunningMode.VIDEO,
    num_hands=2)
detector = vision.HandLandmarker.create_from_options(options)

# STEP 2: Declaração e desenho e das hand connections 
HAND_CONNECTIONS = [
    (0,1),(1,2),(2,3),(3,4),         # polegar  (amarelo)
    (0,5),(5,6),(6,7),(7,8),         # indicador (azul)
    (0,9),(9,10),(10,11),(11,12),    # médio     (verde)
    (0,13),(13,14),(14,15),(15,16),  # anelar    (rosa)
    (0,17),(17,18),(18,19),(19,20),  # mínimo    (cinza)
    (5,9),(9,13),(13,17)             # palma
]

def draw_landmarks_on_image(rgb_image, detection_result):
    image = np.copy(rgb_image)
    h, w, _ = image.shape  # altura e largura para desnormalizar as coordenadas

    for hand_landmarks in detection_result.hand_landmarks:

        # Converte coordenadas normalizadas (0~1) para pixels
        pontos = [
            (int(lm.x * w), int(lm.y * h))
            for lm in hand_landmarks
        ]

        # Desenha as conexões (linhas entre os pontos)
        for start, end in HAND_CONNECTIONS:
            cv2.line(image, pontos[start], pontos[end], (0, 0, 255), 2)

        # Desenha os pontos (landmarks)
        for ponto in pontos:
            cv2.circle(image, ponto, 5, (255, 255, 0), -1)

    return image

# STEP 3: Abrir a câmera
cap = cv2.VideoCapture(0)
timestamp_ms = 0

with detector:
    while cap.isOpened():
        success, image = cap.read()
        if not success:
            print("Ignoring empty camera frame.")
            continue

        # STEP 4: Converter frame para mp.Image (a API exige RGB)
        rgb_image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)
        mp_image = mp.Image(image_format=mp.ImageFormat.SRGB, data=rgb_image)

        # STEP 5: Detectar (passando o timestamp)
        detection_result = detector.detect_for_video(mp_image, timestamp_ms)
        timestamp_ms += 1

        # STEP 6: Desenhar e exibir
        output_image = draw_landmarks_on_image(rgb_image, detection_result)
        output_rgb_image = cv2.cvtColor(output_image, cv2.COLOR_RGB2BGR)
        
        cv2.imshow("MediaPipe Hands", cv2.flip(output_rgb_image, 1))
        if cv2.waitKey(5) & 0xFF == 27:
            break
cap.release()