main.py

from locale import normalize
from unittest import result
import cv2
import mediapipe as mp
from keras.models import load_model
import numpy as np

cap = cv2.VideoCapture(0)

hands = mp.solutions.hands.Hands(max_num_hands=1)

classes = ['A', 'B', 'C', 'D', 'E']
model = load_model('keras_model.h5')
data = np.ndarray(shape=(1, 224, 224, 3), dtype=np.float32)

while True:
    success, img = cap.read()
    frameRGB = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
    results = hands.process(frameRGB)
    handsPoints = result.multi_hand_landmarks
    h, w, _ = img.shape

    if handsPoints != None:
        for hand in handsPoints:
            x_max = 0
            y_max = 0
            x_min = w
            y_min = h
            for lm in hand.landmark:
                x, y = int(lm.x * w), int(lm.y * h)
                if x > x_max:
                    x_max = x
                if x < x_min:
                    x_min = x
                if y > y_max:
                    y_max = y
                if y < y_min:
                    y_min = y
            cv2.rectangle(img, (x_min-50, y_min-50), (x_max+50, y_max+50), (0,255,0), 2)

            try:
                imgCrop = img[y_min+50:y_max-50, x_min-50:x_max+50]
                imgCrop =cv2.resize(imgCrop, (224, 224))
                imgArray = np.asarray(imgCrop)
                normalized_image_array = (imgArray.astype(np.float32) / 127.0) - 1
                data[0] = normalized_image_array
                production = model.predict(data)
                indexVal = np.argmax(production)

                cv2.putText(img, classes[indexVal], (x_min-50, y_min-65), cv2.FONT_HERSHEY_COMPLEX, 3, (0, 0, 255), 5)
            except:
                continue

    cv2.imshow('Imagem', img)
    cv2.waitKey(1)