190919_har_test_train.py

190919_har_test_train.py

from keras.models import Sequential
from keras.layers import Activation, Dense, Dropout,Conv2D,MaxPooling2D,Flatten,ZeroPadding2D
from keras.utils.np_utils import to_categorical
from keras.optimizers import Adagrad
from keras.optimizers import Adam
import numpy as np
from PIL import Image
import os
from sklearn.model_selection import train_test_split
import tensorflow as tf

# 学習用のデータを作る.
image_list = []
label_list = []

for dir in os.listdir("train"):
    if dir == ".DS_Store":
        continue

    dir1 = "./train/" + dir 
    label = 0

    if dir == "001":   
        label = 0
    elif dir == "002": 
        label = 1
    elif dir == "003": 
        label = 2
    elif dir == "004": 
        label = 3
    elif dir == "005": 
        label = 4

    for file in os.listdir(dir1):
        if file != ".DS_Store":
            label_list.append(label)
            filepath = dir1 + "/" + file
            image = Image.open(filepath)
#            image = Image.open(filepath).resize((14, 14)).convert("RGB")
            data = np.asarray(image)
            image_list.append(data)

image_list = np.array(image_list)
image_list = image_list.astype('float32')
image_list = image_list / 255.0
Y = to_categorical(label_list)

print(Y)
print(image_list)

X_train, X_test, y_train, y_test = train_test_split(image_list, Y, test_size=0.20)


# モデルを生成してニューラルネットを構築
model = Sequential()
input_shape=(14, 14, 3)
model.add(ZeroPadding2D(padding=((1, 1), (1, 1)), input_shape=input_shape))
model.add(Conv2D(32, (3, 3),input_shape=input_shape))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(2, 2)))

model.add(ZeroPadding2D(padding=((1, 1), (1, 1)), input_shape=input_shape))
model.add(Conv2D(32, (3, 3)))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(2, 2)))

model.add(Dropout(0.25))

model.add(Flatten())
model.add(Dense(32, activation='relu'))
#model.add(Dense(512))
model.add(Dropout(0.5))
model.add(Dense(5,activation='softmax'))

opt = Adam(lr=0.001)
model.compile(loss="categorical_crossentropy", optimizer=opt, metrics=["accuracy"])
history=model.fit(X_train, y_train, nb_epoch=100)

#print(image_list)

from matplotlib import pyplot as plt

def plot_graph(history):
	epochs = range(len(history.history['acc']))

	plt.plot(epochs,history.history['acc'], marker='.', label='acc')
#	plt.plot(epochs,history.history['val_acc'], marker='.', label='val_acc')
	plt.autoscale()
	plt.title('model accuracy&loss')
	plt.grid()
	plt.xlabel('epoch')
	plt.ylabel('accuracy')
	plt.legend(loc='best')
#	plt.savefig('./acc_graph.png')
#	plt.show()
#	plt.clf()
	plt.plot(epochs,history.history['loss'], marker='.', label='loss')
#	plt.plot(epochs,history.history['val_loss'], marker='.', label='val_loss')
	plt.autoscale()
#	plt.title('model loss')
	plt.grid()
	plt.xlabel('epoch')
	plt.ylabel('loss')
	plt.legend(loc='best')
	plt.savefig('./loss_graph.png')
#	plt.show()
	
plot_graph(history)


model.save('my_mbnet.h5')

converter = tf.lite.TFLiteConverter.from_keras_model_file('my_mbnet.h5')
tflite_model = converter.convert()
open('my_mbnet.tflite', "wb").write(tflite_model)

import subprocess
subprocess.run(['./ncc/ncc','my_mbnet.tflite','my_mbnet.kmodel','-i','tflite','-o',
'k210model','--dataset','images'])