import var5 import numpy as np from tensorflow.keras.models import Model from tensorflow.keras.layers import Input, Convolution2D, MaxPooling2D, Dense, Flatten import matplotlib.pyplot as plt from sklearn.preprocessing import LabelEncoder # Подготовка данных data, labels = var5.gen_data() # генерация данных sz = data.shape[0] img_size = data.shape[1] labels = labels.reshape(sz) encoder = LabelEncoder() # кодирование меток encoder.fit(labels) labels = encoder.transform(labels).reshape(sz, 1) data = data.reshape(sz, img_size**2) data = np.hstack((data, labels)) rng = np.random.default_rng() rng.shuffle(data) # перемешивание данных labels = data[:, -1].reshape(sz, 1) data = data[:, :data.shape[1]-1].reshape(sz, img_size, img_size, 1) data /= np.max(data) # нормализация tr_sz = int(sz*0.9 // 1) # разбиение на тренировочное и тестовое множества train_data = data[:tr_sz, :] train_labels = labels[:tr_sz, :] test_data = data[tr_sz:, :] test_labels = labels[tr_sz:, :] # Задание параметров batch_size = 10
from tensorflow.keras.layers import Dense, Dropout, Flatten
from tensorflow.keras.layers import Conv2D, MaxPooling2D
from sklearn.preprocessing import LabelEncoder
import numpy as np
import matplotlib.pyplot as plt
from var5 import gen_data
def build_model():
return model
size = 3000
test_count = size // 5
data, labels = gen_data(size)
labels = np.asarray(labels).flatten()
encoder = LabelEncoder()
encoder.fit(labels)
labels = encoder.transform(labels)
temp = list(zip(data, labels))
np.random.shuffle(temp)
data, labels = zip(*temp)
data = np.asarray(data).reshape(size, 50, 50, 1)
labels = np.asarray(labels).flatten()
test_data = data[:test_count]
test_labels = labels[:test_count]
train_data = data[test_count:size]
import var5
batch_size = 32
num_epochs = 200
kernel_size = 4
pool_size = 2
conv_depth_1 = 32
conv_depth_2 = 64
drop_prob_1 = 0.25
drop_prob_2 = 0.5
hidden_size = 512
size = 200
img_size = 50
data, labels = var5.gen_data(size, img_size)
labels.reshape(labels.size)
num_classes = np.unique(labels).shape[0]
le.fit(np.unique(labels))
labels = le.transform(labels)
labels = utils.to_categorical(labels, num_classes) # One-hot encode the labels
inp = Input(shape=(img_size, img_size, 1))
# Conv [32] -> Conv [32] -> Pool (with dropout on the pooling layer)
layer = Convolution2D(conv_depth_1, (kernel_size, kernel_size),
padding='same',
activation='relu')(inp)
layer = Convolution2D(conv_depth_1, (kernel_size, kernel_size),
padding='same',
activation='relu')(layer)
from sklearn.utils import shuffle from sklearn.preprocessing import LabelEncoder from tensorflow.keras.models import Sequential from keras.layers import Convolution2D, MaxPooling2D, Dense, Dropout, Flatten import numpy as np from matplotlib import pyplot as plt from var5 import gen_data [data, labels] = gen_data(1000) [data, labels] = shuffle(data, labels) data /= np.max(data) encoder = LabelEncoder() encoder.fit(labels.ravel()) labels = encoder.transform(labels.ravel()) data_len, height, width = data.shape validation_ratio = 0.2 test_ratio = 0.2 train_data = data[: int(data_len*(1 - validation_ratio - test_ratio))] train_labels = labels[: int(data_len*(1 - validation_ratio - test_ratio))] validation_data = data[int(data_len*(1 - validation_ratio - test_ratio)): int(data_len*(1 - test_ratio))] validation_labels = labels[int(data_len*(1 - validation_ratio - test_ratio)): int(data_len*(1 - test_ratio))] test_data = data[int(data_len*(1 - test_ratio)):]
self.table["ind"].append(min_acc_ind)
self.table["class"].append(int(min_acc_class))
self.table["acc"].append(min_acc)
self.table["loss"].append(min_acc_loss)
def on_train_end(self, logs=None):
df = pd.DataFrame(data=self.table)
df.to_csv("LowestAccuracyLog.csv")
def accuracy(y_true, y_pred):
return 1 - abs(y_pred - y_true)
# Подготовка данных
data, labels = var5.gen_data()
sz = data.shape[0]
img_size = data.shape[1]
labels = labels.reshape(sz)
encoder = LabelEncoder()
encoder.fit(labels)
labels = encoder.transform(labels).reshape(sz, 1)
data = data.reshape(sz, img_size**2)
data = np.hstack((data, labels))
rng = np.random.default_rng()
rng.shuffle(data)
labels = data[:, -1].reshape(sz, 1)
data = data[:, :data.shape[1] - 1].reshape(sz, img_size, img_size, 1)
data /= np.max(data)
tr_sz = int(sz * 0.9)
train_data = data[:tr_sz, :]
# Подключение модулей
import numpy as np # Предоставляет общие математические и числовые операции
from sklearn.preprocessing import LabelEncoder # Для преобразования категорий в понятные модели числовые данные
from tensorflow.keras.models import Sequential # Для создания простых моделей используют Sequential,
# При использовании Sequential достаточно добавить несколько своих слоев
from sklearn.model_selection import train_test_split # Деление выборки на тренировочную и тестовую часть
from tensorflow.keras.layers import Dense, Dropout, Flatten, Conv2D
from var5 import gen_data
# Загрузка данных
size = 13000
dataset, labels = gen_data(
size) # dataset - входные данные, labels - выходные данные
dataset = np.asarray(dataset)
labels = np.asarray(labels)
dataset_train, dataset_test, labels_train, labels_test = train_test_split(
dataset, labels, test_size=0.2)
dataset_train = dataset_train.reshape(dataset_train.shape[0], 50, 50, 1)
dataset_test = dataset_test.reshape(dataset_test.shape[0], 50, 50, 1)
dataset_train = dataset_train.astype('float32')
dataset_test = dataset_test.astype('float32')
# Нормализация
dataset_train /= 255
dataset_test /= 255
# Переход от текстовых меток к категориальному вектору (Вектор -> матрица)
encoder = LabelEncoder()
encoder.fit(