Raspberry Pi 深層学習でリアルタイム顔認識（Keras・Open CV）

face_checker_cv_notalk.py

import face_keras as face
import sys, os
from keras.preprocessing.image import load_img, img_to_array
import numpy as np
import cv2
import time

cascade_path = "haarcascade_frontalface_alt.xml"
cascade = cv2.CascadeClassifier(cascade_path)
cam = cv2.VideoCapture(0)
color = (255, 255, 255)

image_size = 32
categories = ["Samuel", "Travolta", "PonDad", "Madsen", "Pam", "Mikami"]

def main():

    while(True):
        ret, frame = cam.read()
        facerect = cascade.detectMultiScale(frame, scaleFactor=1.2, minNeighbors=2, minSize=(10, 10))
        cv2.imwrite("frontalface.png", frame)
        img = cv2.imread("frontalface.png")
        for rect in facerect:
            cv2.rectangle(frame, tuple(rect[0:2]),tuple(rect[0:2] + rect[2:4]), color, thickness=2)
            x = rect[0]
            y = rect[1]
            width = rect[2]
            height = rect[3]
            dst = img[y:y+height, x:x+width]
            cv2.imwrite("output.png", dst)
            cv2.imread("output.png")
            X = []

            img = load_img("./output.png", target_size=(image_size,image_size))
            in_data = img_to_array(img)

            X.append(in_data)
            X = np.array(X)
            X  = X.astype("float")  / 256

            model = face.build_model(X.shape[1:])
            model.load_weights("./image/face-model.h5")

            pre = model.predict(X)
            print(pre)
            if pre[0][0] > 0.9:
                print(categories[0])
                text = categories[0] + str(pre[0][0]*100) + "%"
                font = cv2.FONT_HERSHEY_PLAIN
                cv2.putText(frame,text,(rect[0],rect[1]-10),font, 2, color, 2, cv2.LINE_AA)
            elif pre[0][1] > 0.9:
                print(categories[1])
                text = categories[1] + str(pre[0][1]*100) + "%"
                font = cv2.FONT_HERSHEY_PLAIN
                cv2.putText(frame,text,(rect[0],rect[1]-10),font, 2, color, 2, cv2.LINE_AA)
            elif pre[0][2] > 0.9:
                print(categories[2])
                text = categories[2] + str(pre[0][2]*100) + "%"
                font = cv2.FONT_HERSHEY_PLAIN
                cv2.putText(frame,text,(rect[0],rect[1]-10),font, 2, color, 2, cv2.LINE_AA)

        cv2.imshow("Show FLAME Image", frame)
        time.sleep(0.4)

        k = cv2.waitKey(1)

        if k == ord('q'):
            break

    cam.release()
    cv2.destroyAllWindows()

if __name__ == '__main__':
    main()

face_keras.py

from keras.models import Sequential
from keras.layers import Dense, Dropout, Activation, Flatten
from keras.layers import Convolution2D, MaxPooling2D
from keras.utils import np_utils
import numpy as np

root_dir = "./image/"
categories = ["samuel", "travolta", "pon_dad", "madsen", "pam", "mikami"]
nb_classes = len(categories)
image_size = 32

def main():
    X_train, X_test, y_train, y_test = np.load("./image/face.npy")
    X_train = X_train.astype("float") / 256
    X_test  = X_test.astype("float")  / 256
    y_train = np_utils.to_categorical(y_train, nb_classes)
    y_test  = np_utils.to_categorical(y_test, nb_classes)
    model = model_train(X_train, y_train)
    model_eval(model, X_test, y_test)

def build_model(in_shape):
    model = Sequential()
    model.add(Convolution2D(32, 3, 3,
	border_mode='same',
	input_shape=in_shape))
    model.add(Activation('relu'))
    model.add(MaxPooling2D(pool_size=(2, 2)))
    model.add(Dropout(0.25))
    model.add(Convolution2D(64, 3, 3, border_mode='same'))
    model.add(Activation('relu'))
    model.add(Convolution2D(64, 3, 3))
    model.add(MaxPooling2D(pool_size=(2, 2)))
    model.add(Dropout(0.25))
    model.add(Flatten())
    model.add(Dense(512))
    model.add(Activation('relu'))
    model.add(Dropout(0.5))
    model.add(Dense(nb_classes))
    model.add(Activation('softmax'))
    model.compile(loss='binary_crossentropy',
	optimizer='rmsprop',
	metrics=['accuracy'])
    return model

def model_train(X, y):
    model = build_model(X.shape[1:])
    history = model.fit(X, y, batch_size=32, nb_epoch=30, validation_split=0.1)
    hdf5_file = "./image/face-model.h5"
    model.save_weights(hdf5_file)
    return model

def model_eval(model, X, y):
    score = model.evaluate(X, y)
    print('loss=', score[0])
    print('accuracy=', score[1])

if __name__ == "__main__":
    main()

face_makedata.py

from sklearn import cross_validation
from keras.preprocessing.image import load_img, img_to_array
import os, glob
import numpy as np

root_dir = "./image/"
categories = ["samuel", "travolta", "pon_dad", "madsen", "pam", "mikami"]
nb_classes = len(categories)
image_size = 32

X = []
Y = []
for idx, cat in enumerate(categories):
    files = glob.glob(root_dir + "/" + cat + "/*")
    print("---", cat, "を処理中")
    for i, f in enumerate(files):
        img = load_img(f, target_size=(image_size,image_size))
        data = img_to_array(img)
        X.append(data)
        Y.append(idx)
X = np.array(X)
Y = np.array(Y)

X_train, X_test, y_train, y_test = \
    cross_validation.train_test_split(X, Y)
xy = (X_train, X_test, y_train, y_test)
np.save("./image/face.npy", xy)
print("ok,", len(Y))