Вы не можете выбрать более 25 тем
Темы должны начинаться с буквы или цифры, могут содержать дефисы(-) и должны содержать не более 35 символов.
1 строка
52 KiB
Plaintext
1 строка
52 KiB
Plaintext
{"nbformat":4,"nbformat_minor":0,"metadata":{"colab":{"provenance":[],"gpuType":"T4"},"kernelspec":{"name":"python3","display_name":"Python 3"},"language_info":{"name":"python"},"accelerator":"GPU"},"cells":[{"cell_type":"code","source":[],"metadata":{"id":"1TFPbD2koKmu"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["from google.colab import drive\n","drive.mount('/content/drive')"],"metadata":{"id":"RtZRG8oCnmmC"},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":["1) В среде Google Colab создать новый блокнот (notebook). Импортировать\n","необходимые для работы библиотеки и модули.\n"],"metadata":{"id":"5oURju7duPmk"}},{"cell_type":"code","source":["from tensorflow import keras\n","import matplotlib.pyplot as plt\n","import numpy as np\n","import sklearn\n","from keras.datasets import mnist\n","from sklearn.model_selection import train_test_split\n","from tensorflow.keras.utils import to_categorical\n","from keras.models import Sequential\n","from keras.layers import Dense\n","from PIL import Image"],"metadata":{"id":"09gFd_-eo7eW"},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":["2) Загрузить набор данных MNIST, содержащий размеченные изображения\n","рукописных цифр.\n"],"metadata":{"id":"mheOTPt8uTD2"}},{"cell_type":"code","source":["# загрузка датасета\n","(X_train, y_train), (X_test, y_test) = mnist.load_data()"],"metadata":{"id":"9k2EoBu8pKS0"},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":["3) Разбить набор данных на обучающие и тестовые данные в соотношении\n","60 000:10 000 элементов. При разбиении параметр random_state выбрать\n","равным (4k – 1), где k – номер бригады. Вывести размерности полученных\n","обучающих и тестовых массивов данных.\n","k = 10"],"metadata":{"id":"foMRRdZ-uXOa"}},{"cell_type":"code","source":["# объединяем в один набор\n","X = np.concatenate((X_train, X_test))\n","y = np.concatenate((y_train, y_test))\n","# разбиваем по вариантам\n","X_train, X_test, y_train, y_test = train_test_split(X, y,\n"," test_size = 10000,\n"," train_size = 60000,\n"," random_state = 4*10 - 1)"],"metadata":{"id":"lmfDKm9yptVY"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["# вывод размерностей\n","print('Shape of X train:', X_train.shape)\n","print('Shape of y train:', y_train.shape)"],"metadata":{"id":"ycwVxkDzqYoR"},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":["4) Вывести первые 4 элемента обучающих данных (изображения и метки\n","цифр)."],"metadata":{"id":"8qespHFRugQX"}},{"cell_type":"code","source":["# вывод изображения\n","fig, axes = plt.subplots(1, 4, figsize=(10, 3))\n","for i in range(4):\n"," axes[i].imshow(X_train[i], cmap=plt.get_cmap('gray'))\n"," axes[i].set_title(y_train[i])\n"," axes[i].axis('off')\n","plt.show()"],"metadata":{"id":"HLpjjwweqg_D"},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":["5) Провести предобработку данных: привести обучающие и тестовые данные\n","к формату, пригодному для обучения нейронной сети. Входные данные\n","должны принимать значения от 0 до 1, метки цифр должны быть\n","закодированы по принципу «one-hot encoding». Вывести размерности\n","предобработанных обучающих и тестовых массивов данных."],"metadata":{"id":"4rTauf-dunEM"}},{"cell_type":"code","source":["# развернем каждое изображение 28*28 в вектор 784\n","num_pixels = X_train.shape[1] * X_train.shape[2]\n","X_train = X_train.reshape(X_train.shape[0], num_pixels) / 255\n","X_test = X_test.reshape(X_test.shape[0], num_pixels) / 255\n","print('Shape of transformed X train:', X_train.shape)\n"],"metadata":{"id":"JJlSBKttun8W"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["# переведем метки в one-hot\n","y_train = to_categorical(y_train)\n","y_test = to_categorical(y_test)\n","print('Shape of transformed y train:', y_train.shape)\n","num_classes = y_train.shape[1]\n"],"metadata":{"id":"Q5LqZcAdvhbl"},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":["6) Реализовать модель однослойной нейронной сети и обучить ее на\n","обучающих данных с выделением части обучающих данных в качестве\n","валидационных. Вывести информацию об архитектуре нейронной сети.\n","Вывести график функции ошибки на обучающих и валидационных данных\n","по эпохам."],"metadata":{"id":"NYYbUgM6xzEW"}},{"cell_type":"code","source":["model = Sequential()\n","model.add(Dense(units=num_classes,input_dim=num_pixels, activation='softmax'))\n","model.compile(loss='categorical_crossentropy', optimizer='sgd', metrics=['accuracy'])\n","print(model.summary())"],"metadata":{"id":"IS8-6AUexljX"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["H1 = model.fit(X_train, y_train, batch_size=256, validation_split=0.1, epochs=150)\n"],"metadata":{"id":"IKRNkIbCC24a"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["# вывод графика ошибки по эпохам\n","plt.plot(H1.history['loss'])\n","plt.plot(H1.history['val_loss'])\n","plt.grid()\n","plt.xlabel('Epochs')\n","plt.ylabel('loss')\n","plt.legend(['train_loss', 'val_loss'])\n","plt.title('Loss by epochs')\n","plt.show()"],"metadata":{"id":"kyd2Bh_hE9VO"},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":["7) Применить обученную модель к тестовым данным. Вывести значение\n","функции ошибки и значение метрики качества классификации на тестовых\n","данных."],"metadata":{"id":"Z2QVCB9NEzpM"}},{"cell_type":"code","source":["# Оценка качества работы модели на тестовых данных\n","scores = model.evaluate(X_test, y_test)\n","print('Loss on test data:', scores[0])\n","print('Accuracy on test data:', scores[1])"],"metadata":{"id":"NnPi_avSEuAy"},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":["8) Добавить в модель один скрытый и провести обучение и тестирование\n","(повторить п. 6–7) при 100, 300, 500 нейронах в скрытом слое. По метрике\n","качества классификации на тестовых данных выбрать наилучшее\n","количество нейронов в скрытом слое. В качестве функции активации\n","нейронов в скрытом слое использовать функцию sigmoid."],"metadata":{"id":"ukflwBJgGhu5"}},{"cell_type":"code","source":["model_100 = Sequential ()\n","model_100.add(Dense(units=100,input_dim=num_pixels, activation='sigmoid'))\n","model_100.add(Dense(units=num_classes, activation='softmax'))\n","model_100.compile(loss='categorical_crossentropy', optimizer='sgd', metrics=['accuracy'])\n","print(model_100.summary())"],"metadata":{"id":"VvgnJtzaGi-V"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["H_100 = model_100.fit(X_train, y_train, batch_size=256, validation_split=0.1, epochs=150)"],"metadata":{"id":"foBFtG3hHNsM"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["# вывод графика ошибки по эпохам\n","plt.plot(H_100.history['loss'])\n","plt.plot(H_100.history['val_loss'])\n","plt.grid()\n","plt.xlabel('Epochs')\n","plt.ylabel('loss')\n","plt.legend(['train_loss', 'val_loss'])\n","plt.title('Loss by epochs')\n","plt.show()"],"metadata":{"id":"zHA_blqiJm7-"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["# Оценка качества работы модели на тестовых данных\n","scores = model_100.evaluate(X_test, y_test)\n","print('Loss on test data:', scores[0])\n","print('Accuracy on test data:', scores[1])"],"metadata":{"id":"rxwTYECGJsAi"},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":["300"],"metadata":{"id":"gcDXa-faJ-Xp"}},{"cell_type":"code","source":["model_300 = Sequential ()\n","model_300.add(Dense(units=300,input_dim=num_pixels, activation='sigmoid'))\n","model_300.add(Dense(units=num_classes, activation='softmax'))\n","model_300.compile(loss='categorical_crossentropy', optimizer='sgd', metrics=['accuracy'])\n","print(model_300.summary())"],"metadata":{"id":"CcIlDKU4J5Fy"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["H_300 = model_300.fit(X_train, y_train, batch_size=256, validation_split=0.1, epochs=150)"],"metadata":{"id":"5dChQjjIJ5Fz"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["# вывод графика ошибки по эпохам\n","plt.plot(H_300.history['loss'])\n","plt.plot(H_300.history['val_loss'])\n","plt.grid()\n","plt.xlabel('Epochs')\n","plt.ylabel('loss')\n","plt.legend(['train_loss', 'val_loss'])\n","plt.title('Loss by epochs')\n","plt.show()"],"metadata":{"id":"iTj6TcVzJ5Fz"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["# Оценка качества работы модели на тестовых данных\n","scores = model_300.evaluate(X_test, y_test)\n","print('Loss on test data:', scores[0])\n","print('Accuracy on test data:', scores[1])"],"metadata":{"id":"AaTfpYqZJ5F0"},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":["500"],"metadata":{"id":"ROIhv1chKtlU"}},{"cell_type":"code","source":["model_500 = Sequential ()\n","model_500.add(Dense(units=500,input_dim=num_pixels, activation='sigmoid'))\n","model_500.add(Dense(units=num_classes, activation='softmax'))\n","model_500.compile(loss='categorical_crossentropy', optimizer='sgd', metrics=['accuracy'])\n","print(model_500.summary())"],"metadata":{"id":"APtFO2PMKu6W"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["H_500 = model_500.fit(X_train, y_train, batch_size=256, validation_split=0.1, epochs=150)"],"metadata":{"id":"vC0VRA8yKu6W"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["# вывод графика ошибки по эпохам\n","plt.plot(H_500.history['loss'])\n","plt.plot(H_500.history['val_loss'])\n","plt.grid()\n","plt.xlabel('Epochs')\n","plt.ylabel('loss')\n","plt.legend(['train_loss', 'val_loss'])\n","plt.title('Loss by epochs')\n","plt.show()"],"metadata":{"id":"yEM-YlO1Ku6X"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["# Оценка качества работы модели на тестовых данных\n","scores = model_500.evaluate(X_test, y_test)\n","print('Loss on test data:', scores[0])\n","print('Accuracy on test data:', scores[1])"],"metadata":{"id":"wtNPbUklKu6X"},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":["9) Добавить в наилучшую архитектуру, определенную в п. 8, второй скрытый\n","слой и провести обучение и тестирование (повторить п. 6–7) при 50 и 100\n","нейронах во втором скрытом слое. В качестве функции активации\n","нейронов в скрытом слое использовать функцию sigmoid."],"metadata":{"id":"oxK9QRg7Muj9"}},{"cell_type":"code","source":["model_100_50 = Sequential()\n","model_100_50.add(Dense(units=100, input_dim=num_pixels, activation='sigmoid'))\n","model_100_50.add(Dense(units=50, activation='sigmoid'))\n","model_100_50.add(Dense(units=num_classes, activation='softmax'))\n","model_100_50.compile(loss='categorical_crossentropy', optimizer='sgd', metrics=['accuracy'])\n","print(model_100_50.summary())"],"metadata":{"id":"_nawrX4iNBHW"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["H_100_50 = model_100_50.fit(X_train, y_train, batch_size=256, validation_split=0.1, epochs=150)"],"metadata":{"id":"sAb7w7eINBHX"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["# вывод графика ошибки по эпохам\n","plt.plot(H_100_50.history['loss'])\n","plt.plot(H_100_50.history['val_loss'])\n","plt.grid()\n","plt.xlabel('Epochs')\n","plt.ylabel('loss')\n","plt.legend(['train_loss', 'val_loss'])\n","plt.title('Loss by epochs')\n","plt.show()"],"metadata":{"id":"xXFa8Ob6NBHX"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["# Оценка качества работы модели на тестовых данных\n","scores = model_100_50.evaluate(X_test, y_test)\n","print('Loss on test data:', scores[0])\n","print('Accuracy on test data:', scores[1])"],"metadata":{"id":"CoZo6w5qNBHY"},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":["100 на втрором слое"],"metadata":{"id":"9lx5dZCGN5rS"}},{"cell_type":"code","source":["model_100_100 = Sequential ()\n","model_100_100.add(Dense(units=100,input_dim=num_pixels, activation='sigmoid'))\n","model_100_100.add(Dense(units=100, activation='sigmoid'))\n","model_100_100.add(Dense(units=num_classes, activation='softmax'))\n","model_100_100.compile(loss='categorical_crossentropy', optimizer='sgd', metrics=['accuracy'])\n","print(model_100_100.summary())"],"metadata":{"id":"bymnTOnJN29P"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["H_100_100 = model_100_100.fit(X_train, y_train, batch_size=256, validation_split=0.1, epochs=150)"],"metadata":{"id":"0SaEjqzoN29Q"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["# вывод графика ошибки по эпохам\n","plt.plot(H_100_100.history['loss'])\n","plt.plot(H_100_100.history['val_loss'])\n","plt.grid()\n","plt.xlabel('Epochs')\n","plt.ylabel('loss')\n","plt.legend(['train_loss', 'val_loss'])\n","plt.title('Loss by epochs')\n","plt.show()"],"metadata":{"id":"HTzEyc5SN29Q"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["# Оценка качества работы модели на тестовых данных\n","scores = model.evaluate(X_test, y_test)\n","print('Loss on test data:', scores[0])\n","print('Accuracy on test data:', scores[1])"],"metadata":{"id":"MfOcXeXhN29R"},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":["Таблица с наилучшими архитектурами сети и вывод по ним."],"metadata":{"id":"3rA05vGIDC4H"}},{"cell_type":"markdown","source":["11) Сохранить наилучшую нейронную сеть на диск. Данную нейронную сеть потребуется загрузить с диска в одной из следующих лабораторных работ."],"metadata":{"id":"mUcd8nnoLjXe"}},{"cell_type":"code","source":["model_100.save(\"/content/drive/MyDrive/Colab Notebooks/best_model_100.keras\")"],"metadata":{"id":"HS3dgfuKLocO"},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":["12) Для нейронной сети наилучшей архитектурывывести два тестовых изображения, истинные метки и результат распознавания изображений."],"metadata":{"id":"iWjIiC3oNUZe"}},{"cell_type":"code","source":["n = 333\n","result = model_100.predict(X_test[n:n+1])\n","print('NNoutput:',result)\n","plt.imshow(X_test[n].reshape(28,28),cmap=plt.get_cmap('gray'))\n","plt.show()\n","print('Realmark:',str(np.argmax(y_test[n])))\n","print('NNanswer:',str(np.argmax(result)))"],"metadata":{"id":"h6EsX0sWNV1Y"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["n = 234\n","result = model_100.predict(X_test[n:n+1])\n","print('NNoutput:',result)\n","plt.imshow(X_test[n].reshape(28,28),cmap=plt.get_cmap('gray'))\n","plt.show()\n","print('Realmark:',str(np.argmax(y_test[n])))\n","print('NNanswer:',str(np.argmax(result)))"],"metadata":{"id":"fA6xMRS7Ny_U"},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":["13\n"],"metadata":{"id":"FzvS0hqHefWX"}},{"cell_type":"code","source":["from keras.models import load_model\n","\n","model_100=load_model('/content/drive/MyDrive/Colab Notebooks/best_model_100.keras')"],"metadata":{"id":"N-z0ErgfkHdv"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["\n","file_data=Image.open('/content/drive/MyDrive/Colab Notebooks/IS_lab_7.png')\n","file_data=file_data.convert('L')\n","test_img=np.array(file_data)\n"],"metadata":{"id":"uQXfnnNEcF46"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["#выводсобственногоизображения\n","plt.imshow(test_img,cmap=plt.get_cmap('gray'))\n","plt.show()\n","#предобработка\n","test_img=test_img/255\n","test_img=test_img.reshape(1,num_pixels)\n","#распознавание\n","result=model_100.predict(test_img)\n","print('I think it\\'s',np.argmax(result))"],"metadata":{"id":"RDl8sPM_vnXn"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["\n","file_data_4=Image.open('/content/drive/MyDrive/Colab Notebooks/IS_lab_4.png')\n","file_data_4=file_data_4.convert('L')\n","test_img_4=np.array(file_data_4)"],"metadata":{"id":"M8XBxYIDv1Ui"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["#выводсобственногоизображения\n","plt.imshow(test_img_4,cmap=plt.get_cmap('gray'))\n","plt.show()\n","#предобработка\n","test_img_4=test_img_4/255\n","test_img_4=test_img_4.reshape(1,num_pixels)\n","#распознавание\n","result=model_100.predict(test_img_4)\n","print('I think it\\'s',np.argmax(result))"],"metadata":{"id":"iBTfB4W8v5Pi"},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":["14. Каждому члену бригады создать копию собственного изображения, отличающуюся от оригинала поворотом на 90 градусов в любую сторону. Сохранить изображения. Загрузить, предобработать и подать на вход обученной нейронной сети измененные изображения. Вывести изображения и результаты распознавания. Сделать выводы по результатам эксперимента."],"metadata":{"id":"-HMUj57ZR1N8"}},{"cell_type":"code","source":["from PIL import Image\n","file_data=Image.open('/content/drive/MyDrive/Colab Notebooks/IS_lab_7_90.png')\n","file_data=file_data.convert('L')\n","test_img=np.array(file_data)\n"],"metadata":{"id":"AGQgEWl1PLIv"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["#выводсобственногоизображения\n","plt.imshow(test_img,cmap=plt.get_cmap('gray'))\n","plt.show()\n","#предобработка\n","test_img=test_img/255\n","test_img=test_img.reshape(1,num_pixels)\n","#распознавание\n","result=model_100.predict(test_img)\n","print('Ithinkit\\'s',np.argmax(result))"],"metadata":{"id":"NkaPb4jtRktK"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["from PIL import Image\n","file_data_4=Image.open('/content/drive/MyDrive/Colab Notebooks/IS_lab_4_90.png')\n","file_data_4=file_data_4.convert('L')\n","test_img_4=np.array(file_data_4)\n"],"metadata":{"id":"osOpZLGtU4Sd"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["#выводсобственногоизображения\n","plt.imshow(test_img_4,cmap=plt.get_cmap('gray'))\n","plt.show()\n","#предобработка\n","test_img_4=test_img_4/255\n","test_img_4=test_img_4.reshape(1,num_pixels)\n","#распознавание\n","result=model_100.predict(test_img_4)\n","print('Ithinkit\\'s',np.argmax(result))"],"metadata":{"id":"hR_dW-lLU_Ga"},"execution_count":null,"outputs":[]}]} |