{"metadata":{"kernelspec":{"display_name":"Python 3","language":"python","name":"python3"},"language_info":{"name":"python","version":"3.6.6","mimetype":"text/x-python","codemirror_mode":{"name":"ipython","version":3},"pygments_lexer":"ipython3","nbconvert_exporter":"python","file_extension":".py"},"kaggle":{"accelerator":"gpu","dataSources":[{"sourceId":5441,"databundleVersionId":38425,"sourceType":"competition"},{"sourceId":9900,"sourceType":"datasetVersion","datasetId":6209}],"dockerImageVersionId":11543,"isInternetEnabled":false,"language":"python","sourceType":"notebook","isGpuEnabled":true}},"nbformat_minor":4,"nbformat":4,"cells":[{"cell_type":"markdown","source":"

Cats or Dogs - Usando Redes Convolucionais

\n\n\n
\n\n\n# Conteúdo\n\n- Introdução \n- Carregar pacotes e definir parâmetros \n- Ler os dados \n- Exploração de dados\n - Distribuição de classes\n - Amostras de imagens\n- Modelo \n - Preparar o modelo \n - Treinar o modelo \n - Acurácia e perda na validação \n - Acurácia na validação por classe \n- Preparar submissão \n- Conclusões\n- Referências\n\n\n","metadata":{"_uuid":"1fad66867637226998d21e199e23f70e4eea955d"}},{"cell_type":"markdown","source":"# Introdução \n\n\n## Conjunto de Dados\n\nA pasta **train** contém **25.000** imagens de **cães** e **gatos**. Cada imagem nesta pasta tem o rótulo como parte do nome do arquivo. A pasta **test** contém **12.500** imagens, nomeadas de acordo com um id numérico. \nPara cada imagem no conjunto de teste, você deve prever uma probabilidade de que a imagem seja de um cão (**1** = **cão**, **0** = **gato**).\n\n\n## Método\n\nPara a solução deste problema, usaremos um modelo pré-treinado, ResNet-50, substituindo apenas a última camada.\n","metadata":{"_uuid":"9ac4b97e33b99764c02e38fb58e0b56a03c7fa5f"}},{"cell_type":"markdown","source":"# Carregando Bibliotecas","metadata":{"_uuid":"f4c5a829054852b17a48ffda50c7e35c605911bb"}},{"cell_type":"code","source":"import os, cv2, random\nimport numpy as np\nimport pandas as pd\nimport seaborn as sns\nimport matplotlib.pyplot as plt\nfrom sklearn.model_selection import train_test_split\nfrom sklearn.metrics import classification_report\nfrom tqdm import tqdm\nfrom random import shuffle \nfrom IPython.display import SVG\nfrom keras.utils.vis_utils import model_to_dot\nfrom keras.utils import plot_model\nfrom tensorflow.python.keras.applications import ResNet50\nfrom tensorflow.python.keras.models import Sequential\nfrom tensorflow.python.keras.layers import Dense, Flatten, GlobalAveragePooling2D\n%matplotlib inline ","metadata":{"_uuid":"8f2839f25d086af736a60e9eeb907d3b93b6e0e5","_cell_guid":"b1076dfc-b9ad-4769-8c92-a6c4dae69d19","execution":{"iopub.status.busy":"2024-02-14T20:51:32.921881Z","iopub.execute_input":"2024-02-14T20:51:32.922146Z","iopub.status.idle":"2024-02-14T20:51:34.247543Z","shell.execute_reply.started":"2024-02-14T20:51:32.922097Z","shell.execute_reply":"2024-02-14T20:51:34.246714Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"## Parâmetros\n\nAqui definimos alguns parâmetros usados no modelo. O tamanho da imagem é **224**. \nAs imagens são armazenadas em duas pastas, **train** e **test**. \nExistem duas classes de imagens: **Cão** e **Gato**. \nUsaremos um subconjunto do conjunto de dados de treinamento (**20.000** imagens). Do conjunto de treinamento, **50%** será usado para treinamento, **50%** para validação. \nSerá utilizado um modelo pré-treinado do **ResNet-50**. \nUm número de **10** épocas será usado para o treinamento.\n \n\n","metadata":{"_uuid":"90c407d19e9f09a44cef1714461a1b5d10f1b0d8"}},{"cell_type":"code","source":"TEST_SIZE = 0.5\nRANDOM_STATE = 2018\nBATCH_SIZE = 64\nNO_EPOCHS = 20\nNUM_CLASSES = 2\nSAMPLE_SIZE = 20000\nPATH = '/kaggle/input/dogs-vs-cats-redux-kernels-edition/'\nTRAIN_FOLDER = './train/'\nTEST_FOLDER = './test/'\nIMG_SIZE = 224\nRESNET_WEIGHTS_PATH = '/kaggle/input/resnet50/resnet50_weights_tf_dim_ordering_tf_kernels_notop.h5'","metadata":{"_uuid":"f265a0cfc1bc6379507962130087e7f212e12f83","execution":{"iopub.status.busy":"2024-02-14T20:51:51.939896Z","iopub.execute_input":"2024-02-14T20:51:51.940201Z","iopub.status.idle":"2024-02-14T20:51:51.945364Z","shell.execute_reply.started":"2024-02-14T20:51:51.940131Z","shell.execute_reply":"2024-02-14T20:51:51.944543Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"# Ler os dados\n\nDefinimos a lista de imagens de treino. \nAjustando o valor de **SAMPLE_SIZE** podemos reduzir/aumentar o tamanho do conjunto de treinamento. \nAtualmente, **SAMPLE_SIZE** está definido para **20.000**.\n\n","metadata":{"_uuid":"a1c725fa996e10190a924b977354d82b767c68bf"}},{"cell_type":"code","source":"train_image_path = os.path.join(PATH, \"train.zip\")\ntest_image_path = os.path.join(PATH, \"test.zip\")","metadata":{"execution":{"iopub.status.busy":"2024-02-14T20:51:53.010061Z","iopub.execute_input":"2024-02-14T20:51:53.010366Z","iopub.status.idle":"2024-02-14T20:51:53.014561Z","shell.execute_reply.started":"2024-02-14T20:51:53.010319Z","shell.execute_reply":"2024-02-14T20:51:53.013734Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"import zipfile\nwith zipfile.ZipFile(train_image_path,\"r\") as z:\n z.extractall(\".\")","metadata":{"execution":{"iopub.status.busy":"2024-02-14T20:51:53.210076Z","iopub.execute_input":"2024-02-14T20:51:53.210395Z","iopub.status.idle":"2024-02-14T20:52:06.593078Z","shell.execute_reply.started":"2024-02-14T20:51:53.210343Z","shell.execute_reply":"2024-02-14T20:52:06.592452Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"with zipfile.ZipFile(test_image_path,\"r\") as z:\n z.extractall(\".\")","metadata":{"execution":{"iopub.status.busy":"2024-02-14T20:52:06.594364Z","iopub.execute_input":"2024-02-14T20:52:06.594650Z","iopub.status.idle":"2024-02-14T20:52:14.691757Z","shell.execute_reply.started":"2024-02-14T20:52:06.594598Z","shell.execute_reply":"2024-02-14T20:52:14.690867Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"train_image_list = os.listdir(\"./train/\")[0:SAMPLE_SIZE]\ntest_image_list = os.listdir(\"./test/\")","metadata":{"_uuid":"12d65488fa58182f965485608965ae36f345f183","execution":{"iopub.status.busy":"2024-02-14T20:52:14.692961Z","iopub.execute_input":"2024-02-14T20:52:14.693234Z","iopub.status.idle":"2024-02-14T20:52:14.720011Z","shell.execute_reply.started":"2024-02-14T20:52:14.693158Z","shell.execute_reply":"2024-02-14T20:52:14.719416Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"Definimos uma função para analisar os nomes das imagens para extrair as primeiras 3 letras dos nomes das imagens, o que dá o rótulo da imagem. Será ou um gato ou um cachorro. Estamos usando um codificador one hot, armazenando [1,0] para **gato** e [0,1] para **cachorro**.\n","metadata":{"_uuid":"729f56cad0895facac99467faec510d3d61156a9"}},{"cell_type":"code","source":"def label_pet_image_one_hot_encoder(img):\n pet = img.split('.')[-3]\n if pet == 'cat': return [1,0]\n elif pet == 'dog': return [0,1]","metadata":{"_uuid":"1df3dd73913a3e63a059edec2496e2db1d570c34","execution":{"iopub.status.busy":"2024-02-14T20:52:14.721580Z","iopub.execute_input":"2024-02-14T20:52:14.721874Z","iopub.status.idle":"2024-02-14T20:52:14.726027Z","shell.execute_reply.started":"2024-02-14T20:52:14.721818Z","shell.execute_reply":"2024-02-14T20:52:14.725074Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"Estamos definindo também uma função para processar os dados (tanto o conjunto de treinamento quanto o de teste).","metadata":{"_uuid":"e62d359a2c5c01658610c85c162da998afcbf14d"}},{"cell_type":"code","source":"def process_data(data_image_list, DATA_FOLDER, isTrain=True):\n data_df = []\n for img in tqdm(data_image_list):\n path = os.path.join(DATA_FOLDER,img)\n if(isTrain):\n label = label_pet_image_one_hot_encoder(img)\n else:\n label = img.split('.')[0]\n img = cv2.imread(path,cv2.IMREAD_COLOR)\n img = cv2.resize(img, (IMG_SIZE,IMG_SIZE))\n data_df.append([np.array(img),np.array(label)])\n shuffle(data_df)\n return data_df","metadata":{"_cell_guid":"79c7e3d0-c299-4dcb-8224-4455121ee9b0","_uuid":"d629ff2d2480ee46fbb7e2d37f6b5fab8052498a","execution":{"iopub.status.busy":"2024-02-14T20:52:14.727080Z","iopub.execute_input":"2024-02-14T20:52:14.727326Z","iopub.status.idle":"2024-02-14T20:52:14.737924Z","shell.execute_reply.started":"2024-02-14T20:52:14.727286Z","shell.execute_reply":"2024-02-14T20:52:14.737039Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"Estamos definindo também uma função para processar os dados (tanto o conjunto de treinamento quanto o de teste).","metadata":{"_uuid":"cd19491249f9dd5bc2fd8bfbdfd124fdcdb0cbe1"}},{"cell_type":"code","source":"def plot_image_list_count(data_image_list):\n labels = []\n for img in data_image_list:\n labels.append(img.split('.')[-3])\n sns.countplot(labels)\n plt.title('Cats and Dogs')\n \nplot_image_list_count(train_image_list)","metadata":{"_uuid":"2901e1126ebe160a5a5c5a03dd07d5b8792a5290","execution":{"iopub.status.busy":"2024-02-14T20:53:27.026554Z","iopub.execute_input":"2024-02-14T20:53:27.026808Z","iopub.status.idle":"2024-02-14T20:53:27.230669Z","shell.execute_reply.started":"2024-02-14T20:53:27.026759Z","shell.execute_reply":"2024-02-14T20:53:27.229671Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"plot_image_list_count(os.listdir(TRAIN_FOLDER))","metadata":{"_uuid":"1f36b71d110658540e3cbf3e435e00810ebb2d19","execution":{"iopub.status.busy":"2024-02-14T20:53:27.232310Z","iopub.execute_input":"2024-02-14T20:53:27.232873Z","iopub.status.idle":"2024-02-14T20:53:27.514980Z","shell.execute_reply.started":"2024-02-14T20:53:27.232808Z","shell.execute_reply":"2024-02-14T20:53:27.513984Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"## Amostras de imagens\n\nVamos representar algumas das imagens. Começamos com uma seleção do conjunto de treinamento. Mostraremos as primeiras 25 imagens do conjunto de treinamento.\n\nPrimeiro, processamos os dados de treinamento, lendo as imagens e criando uma tabela com imagens e rótulos. Se os dados são do conjunto de treinamento, o rótulo é o calculado com a codificação one hot; se os dados são do conjunto de teste, o rót\n","metadata":{"_uuid":"c76a6072f4d272ffc7906d56a3008449a569f45b"}},{"cell_type":"code","source":"train = process_data(train_image_list, TRAIN_FOLDER)","metadata":{"_uuid":"d1b8ee45b627b58a3a5d04bb0c64fb18d5e8f592","execution":{"iopub.status.busy":"2024-02-14T20:52:15.271237Z","iopub.execute_input":"2024-02-14T20:52:15.271864Z","iopub.status.idle":"2024-02-14T20:52:56.332387Z","shell.execute_reply.started":"2024-02-14T20:52:15.271793Z","shell.execute_reply":"2024-02-14T20:52:56.331708Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"Em seguida, plotamos a seleção de imagens.","metadata":{"_uuid":"a3583fc9056b32953bae9b90a3aaab7ff451c025"}},{"cell_type":"code","source":"def show_images(data, isTest=False):\n f, ax = plt.subplots(5,5, figsize=(15,15))\n for i,data in enumerate(data[:25]):\n img_num = data[1]\n img_data = data[0]\n label = np.argmax(img_num)\n if label == 1: \n str_label='Dog'\n elif label == 0: \n str_label='Cat'\n if(isTest):\n str_label=\"None\"\n ax[i//5, i%5].imshow(img_data)\n ax[i//5, i%5].axis('off')\n ax[i//5, i%5].set_title(\"Label: {}\".format(str_label))\n plt.show()\n\nshow_images(train)","metadata":{"_uuid":"8a2b305bce44360f693d678a4b7eb9f27b3a7d36","execution":{"iopub.status.busy":"2024-02-14T20:52:56.333501Z","iopub.execute_input":"2024-02-14T20:52:56.333712Z","iopub.status.idle":"2024-02-14T20:52:58.100638Z","shell.execute_reply.started":"2024-02-14T20:52:56.333675Z","shell.execute_reply":"2024-02-14T20:52:58.098981Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"Vamos também mostrar uma seleção do conjunto de treinamento. Preparamos o conjunto de teste.","metadata":{"_uuid":"7cc51ba41c85ca9a93289362ce93b852ef34be6d"}},{"cell_type":"code","source":"test = process_data(test_image_list, TEST_FOLDER, False)","metadata":{"_uuid":"cb0696a3111e5421b3809f91bcb040f32cbde799","execution":{"iopub.status.busy":"2024-02-14T20:52:58.101900Z","iopub.execute_input":"2024-02-14T20:52:58.102180Z","iopub.status.idle":"2024-02-14T20:53:23.191660Z","shell.execute_reply.started":"2024-02-14T20:52:58.102118Z","shell.execute_reply":"2024-02-14T20:53:23.190821Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"Mostrando um conjunto de imagens de teste","metadata":{"_uuid":"417b99a48d96abddb8bc3895a42520690d31208a"}},{"cell_type":"code","source":"show_images(test,True)","metadata":{"_uuid":"485ff5d66200edf7159027b49d4e60619b8b404a","execution":{"iopub.status.busy":"2024-02-14T20:53:23.192843Z","iopub.execute_input":"2024-02-14T20:53:23.193051Z","iopub.status.idle":"2024-02-14T20:53:24.868076Z","shell.execute_reply.started":"2024-02-14T20:53:23.193014Z","shell.execute_reply":"2024-02-14T20:53:24.867086Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"# Model\n\n## Preparar o modelo\n\nIniciando a preparação dos dados\n\n### Preparar os dados de treino","metadata":{"_uuid":"968f2373e44ccea4585d7191b0afc39a92fdff75"}},{"cell_type":"code","source":"X = np.array([i[0] for i in train]).reshape(-1,IMG_SIZE,IMG_SIZE,3)\ny = np.array([i[1] for i in train])","metadata":{"_uuid":"7651e9ef5ff8074165a5208257c5db81ab83b75f","execution":{"iopub.status.busy":"2024-02-14T20:53:24.869429Z","iopub.execute_input":"2024-02-14T20:53:24.869682Z","iopub.status.idle":"2024-02-14T20:53:27.025123Z","shell.execute_reply.started":"2024-02-14T20:53:24.869630Z","shell.execute_reply":"2024-02-14T20:53:27.024375Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"### Preparar o modelo\n\nInicializamos o modelo **ResNet-50**, adicionando uma última camada adicional do tipo **Dense**, com função de ativação **softmax**.\n\nTambém configuramos a primeira camada do modelo para não ser treinável, porque o modelo **ResNet-50** já foi treinado.","metadata":{"_uuid":"802f6921a5ab7e2b75b6cb239bedd71cb2c0e4cd"}},{"cell_type":"code","source":"from tensorflow.python.keras.applications import ResNet50\nfrom tensorflow.python.keras.models import Sequential\nfrom tensorflow.python.keras.layers import Dense, Flatten, GlobalAveragePooling2D\n\nmodel = Sequential()\nmodel.add(ResNet50(include_top=False, pooling='max', weights=RESNET_WEIGHTS_PATH))\nmodel.add(Dense(NUM_CLASSES, activation='softmax'))\n# ResNet-50 model is already trained, should not be trained\nmodel.layers[0].trainable = True\n","metadata":{"_uuid":"94892424cb4e898d6b7c3df1269cfab6371c4eb9","execution":{"iopub.status.busy":"2024-02-14T20:54:14.691358Z","iopub.execute_input":"2024-02-14T20:54:14.691632Z","iopub.status.idle":"2024-02-14T20:54:27.903868Z","shell.execute_reply.started":"2024-02-14T20:54:14.691591Z","shell.execute_reply":"2024-02-14T20:54:27.903278Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"Compilar o modelo\n\nNós compilamos o modelo, utilizando um otimizador **sigmoid**, a função de perda como **entropia cruzada categórica** e a métrica **precisão**.","metadata":{"_uuid":"cfb6886ef6d14fc9f1cdd1857fdd07f9639e6442"}},{"cell_type":"code","source":"model.compile(optimizer='sgd', loss='categorical_crossentropy', metrics=['accuracy'])","metadata":{"_uuid":"c24a3f51ef3487e15efdc8c2e876aea7c429b876","execution":{"iopub.status.busy":"2024-02-14T20:54:31.116953Z","iopub.execute_input":"2024-02-14T20:54:31.117258Z","iopub.status.idle":"2024-02-14T20:54:31.247689Z","shell.execute_reply.started":"2024-02-14T20:54:31.117205Z","shell.execute_reply":"2024-02-14T20:54:31.247010Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"Resumo do modelo\n\nNós plotamos a descrição do modelo. Podemos ver que o modelo **ResNet-50** representa a 1ª camada do nosso modelo, do tipo **Model**.","metadata":{"_uuid":"52e06decf4d5ea9e099b4669681b8ebb36ccd7bd"}},{"cell_type":"code","source":"model.summary()","metadata":{"_uuid":"3e2a3ff8c2dd896fd9ca725625587564165d0727","execution":{"iopub.status.busy":"2024-02-14T20:54:34.366601Z","iopub.execute_input":"2024-02-14T20:54:34.366878Z","iopub.status.idle":"2024-02-14T20:54:34.380151Z","shell.execute_reply.started":"2024-02-14T20:54:34.366837Z","shell.execute_reply":"2024-02-14T20:54:34.379069Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"Vamos também mostrar a representação gráfica do modelo usando **plot_model**.","metadata":{"_uuid":"6bd39063f81d0b87f80542c6b8ce74035e2de28e"}},{"cell_type":"code","source":"plot_model(model, to_file='model.png')\nSVG(model_to_dot(model).create(prog='dot', format='svg'))","metadata":{"_uuid":"64561287fa8faaa756ba4043a2a3c671810a9dbf","execution":{"iopub.status.busy":"2024-02-14T20:54:38.826243Z","iopub.execute_input":"2024-02-14T20:54:38.826520Z","iopub.status.idle":"2024-02-14T20:54:40.353380Z","shell.execute_reply.started":"2024-02-14T20:54:38.826478Z","shell.execute_reply":"2024-02-14T20:54:40.352329Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"### Dividir os dados de treino em treino e validação\n\nDividimos os dados de treino em duas partes. Uma será reservada para o conjunto de treino, a segunda para o conjunto de validação. Apenas o subconjunto de treino dos dados será usado para treinar o modelo; o conjunto de validação será usado para validação, durante o treino.","metadata":{"_uuid":"cba1cd58d8f3e3240fafe88ba5ac4c8cd3885008"}},{"cell_type":"code","source":"X_train, X_val, y_train, y_val = train_test_split(X, y, test_size=TEST_SIZE, random_state=RANDOM_STATE)","metadata":{"_uuid":"e80c5bd03d275807865c424187f752b521c99f15","execution":{"iopub.status.busy":"2024-02-14T20:54:42.847695Z","iopub.execute_input":"2024-02-14T20:54:42.847980Z","iopub.status.idle":"2024-02-14T20:54:44.812969Z","shell.execute_reply.started":"2024-02-14T20:54:42.847934Z","shell.execute_reply":"2024-02-14T20:54:44.812285Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":" y_train","metadata":{"execution":{"iopub.status.busy":"2024-02-14T20:54:44.814229Z","iopub.execute_input":"2024-02-14T20:54:44.814523Z","iopub.status.idle":"2024-02-14T20:54:44.820145Z","shell.execute_reply.started":"2024-02-14T20:54:44.814469Z","shell.execute_reply":"2024-02-14T20:54:44.819313Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"## Treinar o modelo\n\nAgora estamos prontos para treinar nosso modelo.","metadata":{"_uuid":"cb28fa20f06729cb20a31e7b5c1f99970ebe62fa"}},{"cell_type":"code","source":"train_model = model.fit(X_train, y_train,\n batch_size=BATCH_SIZE,\n epochs=3,\n verbose=1,\n validation_data=(X_val, y_val))","metadata":{"_uuid":"3899d6c135949c4546d5aa10291fb5b1ff4498da","execution":{"iopub.status.busy":"2024-02-14T21:01:23.980929Z","iopub.execute_input":"2024-02-14T21:01:23.981259Z","iopub.status.idle":"2024-02-14T21:06:01.723512Z","shell.execute_reply.started":"2024-02-14T21:01:23.981194Z","shell.execute_reply":"2024-02-14T21:06:01.722634Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"## Acurácia e perda na validação\n\nVamos mostrar a acurácia de treino e validação no mesmo gráfico. Da mesma forma, representaremos a perda de treino e validação no mesmo gráfico.","metadata":{"_uuid":"e9c9159b0a2f2fc8bafe9bfc1983c30a846f099a"}},{"cell_type":"code","source":"def plot_accuracy_and_loss(train_model):\n hist = train_model.history\n acc = hist['acc']\n val_acc = hist['val_acc']\n loss = hist['loss']\n val_loss = hist['val_loss']\n epochs = range(len(acc))\n f, ax = plt.subplots(1,2, figsize=(14,6))\n ax[0].plot(epochs, acc, 'g', label='Training accuracy')\n ax[0].plot(epochs, val_acc, 'r', label='Validation accuracy')\n ax[0].set_title('Training and validation accuracy')\n ax[0].legend()\n ax[1].plot(epochs, loss, 'g', label='Training loss')\n ax[1].plot(epochs, val_loss, 'r', label='Validation loss')\n ax[1].set_title('Training and validation loss')\n ax[1].legend()\n plt.show()\nplot_accuracy_and_loss(train_model)","metadata":{"_uuid":"5a7b3d7b9494a3a1a75db099d224a18b1bf1d8e3","execution":{"iopub.status.busy":"2024-02-14T21:06:01.724712Z","iopub.execute_input":"2024-02-14T21:06:01.724921Z","iopub.status.idle":"2024-02-14T21:06:02.163475Z","shell.execute_reply.started":"2024-02-14T21:06:01.724883Z","shell.execute_reply":"2024-02-14T21:06:02.162445Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"Vamos também mostrar a acurácia e perda na validação numéricas.","metadata":{"_uuid":"130f99193a6b24b40c5a28f84a97b84e6d39784c"}},{"cell_type":"code","source":"score = model.evaluate(X_val, y_val, verbose=0)\nprint('Validation loss:', score[0])\nprint('Validation accuracy:', score[1])","metadata":{"_uuid":"0935a9e239a7af382208ecede99bd2afb5098482","execution":{"iopub.status.busy":"2024-02-14T21:06:02.165379Z","iopub.execute_input":"2024-02-14T21:06:02.165978Z","iopub.status.idle":"2024-02-14T21:06:25.398223Z","shell.execute_reply.started":"2024-02-14T21:06:02.165908Z","shell.execute_reply":"2024-02-14T21:06:25.397322Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"### Test data prediction","metadata":{"_uuid":"2e2199cdb3dbef1620298f0b9060908b7e2d4f90"}},{"cell_type":"code","source":"pred_list = []\nimg_list = []\nfor img in tqdm(test):\n img_data = img[0]\n img_idx = img[1]\n data = img_data.reshape(-1,IMG_SIZE,IMG_SIZE,3)\n predicted = model.predict([data])[0]\n img_list.append(img_idx)\n pred_list.append(predicted[1])","metadata":{"_uuid":"15afbef65cdfe94ffa3a2e63ed97f0fdd9a34d89","execution":{"iopub.status.busy":"2024-02-14T21:07:29.411070Z","iopub.execute_input":"2024-02-14T21:07:29.411379Z","iopub.status.idle":"2024-02-14T21:09:36.571535Z","shell.execute_reply.started":"2024-02-14T21:07:29.411332Z","shell.execute_reply":"2024-02-14T21:09:36.570825Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"","metadata":{},"execution_count":null,"outputs":[]}]}