Raspberry Pi 深層学習で「赤りんご」と「青りんご」を見分ける（Keras・Open CV）

apple_checker_cv.py

import apple_keras as apple
import sys, os
import numpy as np
import subprocess
import cv2
from keras.preprocessing.image import load_img, img_to_array

cam = cv2.VideoCapture(0)
image_size = 32
categories = ["赤りんご", "青りんご"]

def main():

    def jtalk(t):
        open_jtalk=['open_jtalk']
        mech=['-x','/var/lib/mecab/dic/open-jtalk/naist-jdic']
        htsvoice=['-m','/usr/share/hts-voice/mei/mei_happy.htsvoice']
        speed=['-r','1.0']
        outwav=['-ow','open_jtalk.wav']
        cmd=open_jtalk+mech+htsvoice+speed+outwav
        c = subprocess.Popen(cmd,stdin=subprocess.PIPE)
        c.stdin.write(t)
        c.stdin.close()
        c.wait()
        aplay = ['aplay','-q','open_jtalk.wav']
        wr = subprocess.Popen(aplay)

    while(True):
        ret, frame = cam.read()
        cv2.imshow("Show FLAME Image", frame)

        k = cv2.waitKey(1)
        if k == ord('s'):
            cv2.imwrite("output.png", frame)
            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 = apple.build_model(X.shape[1:])
            model.load_weights("./image/apple-model.h5")

            pre = model.predict(X)
            print(pre)
            if pre[0][0] > 0.9:
                print(categories[0])
                text = 'これは' + categories[0]+ 'だよ'
                text = text.encode('utf-8')
                jtalk(text)
            elif pre[0][1] > 0.9:
                print(categories[1])
                text = 'これは' + categories[1]+ 'だよ'
                text = text.encode('utf-8')
                jtalk(text)

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

    cam.release()
    cv2.destroyAllWindows()

if __name__ == '__main__':
    main()

apple_keras.py

# https://github.com/fchollet/keras/blob/master/examples/cifar10_cnn.py

import numpy as np
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

root_dir = "./image/"
categories = ["red_apple", "green_apple"]
nb_classes = len(categories)
image_size = 32

def main():
    X_train, X_test, y_train, y_test = np.load("./image/apple.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=10, validation_split=0.1)
    hdf5_file = "./image/apple-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()

apple_makedata.py

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

root_dir = "./image/"
categories = ["red_apple", "green_apple"]
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/apple.npy", xy)
print("ok,", len(Y))

Licence.txt

The MIT License (MIT)