cropped.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Found 703 files belonging to 2 classes.\n",
      "Using 605 files for training.\n",
      "Found 703 files belonging to 2 classes.\n",
      "Using 98 files for validation.\n"
     ]
    }
   ],
   "source": [
    "import tensorflow as tf \n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "\n",
    "seed = 13\n",
    "batch_size = 16 \n",
    "img_height = 224 \n",
    "img_width = 224 \n",
    "img_size = (img_height, img_width)\n",
    "size = (img_height, img_width)\n",
    "data_dir = 'D:\\Downloads\\cxr_tcc_cropped'\n",
    "weight_path = 'D:\\Downloads\\\\brucechou1983_CheXNet_Keras_0.3.0_weights.h5'\n",
    "best_model_path = 'D:\\Documentos\\projeto convs\\models\\cropped'\n",
    "\n",
    "\n",
    "## Split do dataset.\n",
    "\n",
    "full_train_ds = tf.keras.preprocessing.image_dataset_from_directory(\n",
    "    data_dir,\n",
    "    validation_split=0.14,\n",
    "    subset=\"training\",\n",
    "    seed= seed,\n",
    "    image_size= img_size,\n",
    "    batch_size = batch_size\n",
    ")\n",
    "test_ds = tf.keras.preprocessing.image_dataset_from_directory(\n",
    "    data_dir,\n",
    "    validation_split=0.14, #98 imagens pra teste\n",
    "    subset=\"validation\",\n",
    "    seed= seed,\n",
    "    image_size= img_size,\n",
    "    batch_size = batch_size\n",
    ")\n",
    "\n",
    "full_train_ds.shuffle(1000, reshuffle_each_iteration = False, seed = seed) #embaralha e desativa shuffle pois usar o map que virá depois itera o dataset, o reshuffle pode causar vazamento de dados\n",
    "\n",
    "def recover(x, y):\n",
    "    return y\n",
    "\n",
    "def is_val(x, y):\n",
    "    return x%6 == 0 # 1/6 do dataset ~ 100 imagens pra validação\n",
    "\n",
    "def is_train(x,y):\n",
    "    return not is_val(x,y)\n",
    "\n",
    "val_ds = full_train_ds.enumerate().filter(is_val).map(recover)\n",
    "train_ds = full_train_ds.enumerate().filter(is_train).map(recover)\n",
    "\n",
    "full_train_ds.shuffle(1000, reshuffle_each_iteration = True, seed = seed) #reseta o shuffle para acontecer o shuffle entre épocas do treino\n",
    "\n",
    "\n",
    "import keras\n",
    "from keras import layers\n",
    "import keras_tuner as kt\n",
    "\n",
    "channels = tuple([3])\n",
    "shape = size + channels\n",
    "\n",
    "def get_model(hp):\n",
    "    inputs = keras.Input(shape = shape)\n",
    "    x = layers.RandomRotation(0.05)(inputs) #tentar rotacionar menos\n",
    "    x = layers.Rescaling(scale=1./127.5, offset=-1)(x) #normalização \"manual\"\n",
    "\n",
    "    chexnet = tf.keras.applications.DenseNet121(weights= None,\n",
    "                                                include_top = False, \n",
    "                                                input_tensor = x,\n",
    "                                                input_shape = shape\n",
    "                                               )\n",
    "\n",
    "    out = tf.keras.layers.Dense(14, activation='sigmoid')(chexnet.output)\n",
    "    model = tf.keras.Model(inputs, out)\n",
    "    \n",
    "    for layer in chexnet.layers:\n",
    "        layer.trainable = False\n",
    "    model.load_weights(weight_path)\n",
    "    \n",
    "    x =  layers.GlobalAveragePooling2D()(chexnet.output)\n",
    "    preds = layers.Dense(units = 1, activation = 'sigmoid', name='prediction')(x)\n",
    "    model = tf.keras.Model(model.input, preds)\n",
    "    #Tune LR\n",
    "    hp_lr = hp.Choice('learning_rate', values = [1e-2, 1e-3, 1e-4])\n",
    "    opt = tf.keras.optimizers.SGD(learning_rate = hp_lr)\n",
    "    model.compile(optimizer = opt, loss = 'binary_crossentropy', metrics = ['accuracy'])        \n",
    "    return model\n",
    "\n",
    "tuner = kt.RandomSearch(get_model, objective = kt.Objective('val_accuracy', direction = 'max'), max_trials = 3  , seed = 1337, overwrite = True)\n",
    "\n",
    "from keras.callbacks import ModelCheckpoint, EarlyStopping, ReduceLROnPlateau\n",
    "\n",
    "#Callbacks\n",
    "checkpoint = ModelCheckpoint(best_model_path, monitor='val_loss', verbose=1, \n",
    "                                save_best_only=True, mode='min', save_weights_only = True)\n",
    "\n",
    "reduceLROnPlat = ReduceLROnPlateau(monitor='val_loss', factor=0.5, \n",
    "                                patience=3, \n",
    "                                verbose=1, mode='min', min_delta=0.0001, cooldown=2, min_lr=1e-6)\n",
    "early = EarlyStopping(monitor=\"val_loss\", \n",
    "                    mode=\"min\", \n",
    "                    patience=15) \n",
    "\n",
    "callbacks = [checkpoint, early, reduceLROnPlat]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Trial 3 Complete [00h 11m 09s]\n",
      "val_accuracy: 0.8303571343421936\n",
      "\n",
      "Best val_accuracy So Far: 0.8303571343421936\n",
      "Total elapsed time: 00h 29m 36s\n",
      "INFO:tensorflow:Oracle triggered exit\n",
      "Best learning rate was 0.001.\n"
     ]
    }
   ],
   "source": [
    "tuner.search(train_ds, batch_size = batch_size, epochs = 100, validation_data = val_ds, callbacks = callbacks)\n",
    "best_hps = tuner.get_best_hyperparameters(5)\n",
    "print(f\"Best learning rate was {best_hps[1].get('learning_rate')}.\")\n",
    "model = get_model(best_hps[0])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Full Training"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/100\n",
      "\n",
      "Epoch 00001: val_loss improved from inf to 0.66256, saving model to D:\\Documentos\\projeto convs\\models\\cropped\n",
      "38/38 - 45s - loss: 0.6758 - accuracy: 0.5603 - val_loss: 0.6626 - val_accuracy: 0.5816 - lr: 0.0100 - 45s/epoch - 1s/step\n",
      "Epoch 2/100\n",
      "\n",
      "Epoch 00002: val_loss improved from 0.66256 to 0.64194, saving model to D:\\Documentos\\projeto convs\\models\\cropped\n",
      "38/38 - 9s - loss: 0.6650 - accuracy: 0.6017 - val_loss: 0.6419 - val_accuracy: 0.6224 - lr: 0.0100 - 9s/epoch - 240ms/step\n",
      "Epoch 3/100\n",
      "\n",
      "Epoch 00003: val_loss improved from 0.64194 to 0.62968, saving model to D:\\Documentos\\projeto convs\\models\\cropped\n",
      "38/38 - 9s - loss: 0.6521 - accuracy: 0.6661 - val_loss: 0.6297 - val_accuracy: 0.7041 - lr: 0.0100 - 9s/epoch - 237ms/step\n",
      "Epoch 4/100\n",
      "\n",
      "Epoch 00004: val_loss improved from 0.62968 to 0.61980, saving model to D:\\Documentos\\projeto convs\\models\\cropped\n",
      "38/38 - 9s - loss: 0.6439 - accuracy: 0.6661 - val_loss: 0.6198 - val_accuracy: 0.7347 - lr: 0.0100 - 9s/epoch - 236ms/step\n",
      "Epoch 5/100\n",
      "\n",
      "Epoch 00005: val_loss improved from 0.61980 to 0.60661, saving model to D:\\Documentos\\projeto convs\\models\\cropped\n",
      "38/38 - 9s - loss: 0.6340 - accuracy: 0.6645 - val_loss: 0.6066 - val_accuracy: 0.7347 - lr: 0.0100 - 9s/epoch - 239ms/step\n",
      "Epoch 6/100\n",
      "\n",
      "Epoch 00006: val_loss did not improve from 0.60661\n",
      "38/38 - 8s - loss: 0.6298 - accuracy: 0.6777 - val_loss: 0.6130 - val_accuracy: 0.6837 - lr: 0.0100 - 8s/epoch - 218ms/step\n",
      "Epoch 7/100\n",
      "\n",
      "Epoch 00007: val_loss improved from 0.60661 to 0.59209, saving model to D:\\Documentos\\projeto convs\\models\\cropped\n",
      "38/38 - 9s - loss: 0.6118 - accuracy: 0.7190 - val_loss: 0.5921 - val_accuracy: 0.7245 - lr: 0.0100 - 9s/epoch - 238ms/step\n",
      "Epoch 8/100\n",
      "\n",
      "Epoch 00008: val_loss improved from 0.59209 to 0.59104, saving model to D:\\Documentos\\projeto convs\\models\\cropped\n",
      "38/38 - 9s - loss: 0.6041 - accuracy: 0.6942 - val_loss: 0.5910 - val_accuracy: 0.6939 - lr: 0.0100 - 9s/epoch - 237ms/step\n",
      "Epoch 9/100\n",
      "\n",
      "Epoch 00009: val_loss did not improve from 0.59104\n",
      "38/38 - 8s - loss: 0.6050 - accuracy: 0.7058 - val_loss: 0.6009 - val_accuracy: 0.6531 - lr: 0.0100 - 8s/epoch - 219ms/step\n",
      "Epoch 10/100\n",
      "\n",
      "Epoch 00010: val_loss did not improve from 0.59104\n",
      "38/38 - 8s - loss: 0.5932 - accuracy: 0.7289 - val_loss: 0.5912 - val_accuracy: 0.6837 - lr: 0.0100 - 8s/epoch - 220ms/step\n",
      "Epoch 11/100\n",
      "\n",
      "Epoch 00011: val_loss improved from 0.59104 to 0.56629, saving model to D:\\Documentos\\projeto convs\\models\\cropped\n",
      "38/38 - 9s - loss: 0.5942 - accuracy: 0.6975 - val_loss: 0.5663 - val_accuracy: 0.7347 - lr: 0.0100 - 9s/epoch - 237ms/step\n",
      "Epoch 12/100\n",
      "\n",
      "Epoch 00012: val_loss did not improve from 0.56629\n",
      "38/38 - 8s - loss: 0.5847 - accuracy: 0.7074 - val_loss: 0.5666 - val_accuracy: 0.7041 - lr: 0.0100 - 8s/epoch - 218ms/step\n",
      "Epoch 13/100\n",
      "\n",
      "Epoch 00013: val_loss improved from 0.56629 to 0.55737, saving model to D:\\Documentos\\projeto convs\\models\\cropped\n",
      "38/38 - 9s - loss: 0.5818 - accuracy: 0.7008 - val_loss: 0.5574 - val_accuracy: 0.7449 - lr: 0.0100 - 9s/epoch - 237ms/step\n",
      "Epoch 14/100\n",
      "\n",
      "Epoch 00014: val_loss did not improve from 0.55737\n",
      "38/38 - 8s - loss: 0.5803 - accuracy: 0.6992 - val_loss: 0.5646 - val_accuracy: 0.7143 - lr: 0.0100 - 8s/epoch - 218ms/step\n",
      "Epoch 15/100\n",
      "\n",
      "Epoch 00015: val_loss did not improve from 0.55737\n",
      "38/38 - 8s - loss: 0.5791 - accuracy: 0.7223 - val_loss: 0.5579 - val_accuracy: 0.7041 - lr: 0.0100 - 8s/epoch - 220ms/step\n",
      "Epoch 16/100\n",
      "\n",
      "Epoch 00016: val_loss improved from 0.55737 to 0.55217, saving model to D:\\Documentos\\projeto convs\\models\\cropped\n",
      "38/38 - 9s - loss: 0.5694 - accuracy: 0.7124 - val_loss: 0.5522 - val_accuracy: 0.7041 - lr: 0.0100 - 9s/epoch - 240ms/step\n",
      "Epoch 17/100\n",
      "\n",
      "Epoch 00017: val_loss improved from 0.55217 to 0.54308, saving model to D:\\Documentos\\projeto convs\\models\\cropped\n",
      "38/38 - 9s - loss: 0.5648 - accuracy: 0.7306 - val_loss: 0.5431 - val_accuracy: 0.7347 - lr: 0.0100 - 9s/epoch - 241ms/step\n",
      "Epoch 18/100\n",
      "\n",
      "Epoch 00018: val_loss did not improve from 0.54308\n",
      "38/38 - 8s - loss: 0.5565 - accuracy: 0.7372 - val_loss: 0.5486 - val_accuracy: 0.7041 - lr: 0.0100 - 8s/epoch - 218ms/step\n",
      "Epoch 19/100\n",
      "\n",
      "Epoch 00019: val_loss did not improve from 0.54308\n",
      "38/38 - 8s - loss: 0.5712 - accuracy: 0.7322 - val_loss: 0.5463 - val_accuracy: 0.7041 - lr: 0.0100 - 8s/epoch - 216ms/step\n",
      "Epoch 20/100\n",
      "\n",
      "Epoch 00020: val_loss improved from 0.54308 to 0.53705, saving model to D:\\Documentos\\projeto convs\\models\\cropped\n",
      "38/38 - 9s - loss: 0.5629 - accuracy: 0.7124 - val_loss: 0.5371 - val_accuracy: 0.7143 - lr: 0.0100 - 9s/epoch - 237ms/step\n",
      "Epoch 21/100\n",
      "\n",
      "Epoch 00021: val_loss did not improve from 0.53705\n",
      "38/38 - 8s - loss: 0.5663 - accuracy: 0.7074 - val_loss: 0.5377 - val_accuracy: 0.6837 - lr: 0.0100 - 8s/epoch - 219ms/step\n",
      "Epoch 22/100\n",
      "\n",
      "Epoch 00022: val_loss improved from 0.53705 to 0.53183, saving model to D:\\Documentos\\projeto convs\\models\\cropped\n",
      "38/38 - 9s - loss: 0.5718 - accuracy: 0.7124 - val_loss: 0.5318 - val_accuracy: 0.7245 - lr: 0.0100 - 9s/epoch - 238ms/step\n",
      "Epoch 23/100\n",
      "\n",
      "Epoch 00023: val_loss did not improve from 0.53183\n",
      "38/38 - 8s - loss: 0.5547 - accuracy: 0.7207 - val_loss: 0.5329 - val_accuracy: 0.7143 - lr: 0.0100 - 8s/epoch - 220ms/step\n",
      "Epoch 24/100\n",
      "\n",
      "Epoch 00024: val_loss did not improve from 0.53183\n",
      "38/38 - 8s - loss: 0.5519 - accuracy: 0.7091 - val_loss: 0.5399 - val_accuracy: 0.7143 - lr: 0.0100 - 8s/epoch - 221ms/step\n",
      "Epoch 25/100\n",
      "\n",
      "Epoch 00025: val_loss did not improve from 0.53183\n",
      "\n",
      "Epoch 00025: ReduceLROnPlateau reducing learning rate to 0.004999999888241291.\n",
      "38/38 - 8s - loss: 0.5453 - accuracy: 0.7537 - val_loss: 0.5400 - val_accuracy: 0.7143 - lr: 0.0100 - 8s/epoch - 222ms/step\n",
      "Epoch 26/100\n",
      "\n",
      "Epoch 00026: val_loss improved from 0.53183 to 0.53108, saving model to D:\\Documentos\\projeto convs\\models\\cropped\n",
      "38/38 - 9s - loss: 0.5453 - accuracy: 0.7223 - val_loss: 0.5311 - val_accuracy: 0.7041 - lr: 0.0050 - 9s/epoch - 237ms/step\n",
      "Epoch 27/100\n",
      "\n",
      "Epoch 00027: val_loss did not improve from 0.53108\n",
      "38/38 - 8s - loss: 0.5546 - accuracy: 0.7207 - val_loss: 0.5366 - val_accuracy: 0.7143 - lr: 0.0050 - 8s/epoch - 219ms/step\n",
      "Epoch 28/100\n",
      "\n",
      "Epoch 00028: val_loss improved from 0.53108 to 0.52733, saving model to D:\\Documentos\\projeto convs\\models\\cropped\n",
      "38/38 - 9s - loss: 0.5484 - accuracy: 0.7223 - val_loss: 0.5273 - val_accuracy: 0.6939 - lr: 0.0050 - 9s/epoch - 239ms/step\n",
      "Epoch 29/100\n",
      "\n",
      "Epoch 00029: val_loss did not improve from 0.52733\n",
      "38/38 - 8s - loss: 0.5507 - accuracy: 0.7223 - val_loss: 0.5406 - val_accuracy: 0.7143 - lr: 0.0050 - 8s/epoch - 223ms/step\n",
      "Epoch 30/100\n",
      "\n",
      "Epoch 00030: val_loss did not improve from 0.52733\n",
      "38/38 - 8s - loss: 0.5309 - accuracy: 0.7636 - val_loss: 0.5308 - val_accuracy: 0.7245 - lr: 0.0050 - 8s/epoch - 218ms/step\n",
      "Epoch 31/100\n",
      "\n",
      "Epoch 00031: val_loss improved from 0.52733 to 0.52571, saving model to D:\\Documentos\\projeto convs\\models\\cropped\n",
      "38/38 - 9s - loss: 0.5484 - accuracy: 0.7455 - val_loss: 0.5257 - val_accuracy: 0.7041 - lr: 0.0050 - 9s/epoch - 238ms/step\n",
      "Epoch 32/100\n",
      "\n",
      "Epoch 00032: val_loss did not improve from 0.52571\n",
      "38/38 - 8s - loss: 0.5381 - accuracy: 0.7587 - val_loss: 0.5264 - val_accuracy: 0.7041 - lr: 0.0050 - 8s/epoch - 219ms/step\n",
      "Epoch 33/100\n",
      "\n",
      "Epoch 00033: val_loss improved from 0.52571 to 0.52172, saving model to D:\\Documentos\\projeto convs\\models\\cropped\n",
      "38/38 - 9s - loss: 0.5282 - accuracy: 0.7570 - val_loss: 0.5217 - val_accuracy: 0.7041 - lr: 0.0050 - 9s/epoch - 237ms/step\n",
      "Epoch 34/100\n",
      "\n",
      "Epoch 00034: val_loss improved from 0.52172 to 0.51850, saving model to D:\\Documentos\\projeto convs\\models\\cropped\n",
      "38/38 - 9s - loss: 0.5437 - accuracy: 0.7289 - val_loss: 0.5185 - val_accuracy: 0.7143 - lr: 0.0050 - 9s/epoch - 237ms/step\n",
      "Epoch 35/100\n",
      "\n",
      "Epoch 00035: val_loss improved from 0.51850 to 0.51810, saving model to D:\\Documentos\\projeto convs\\models\\cropped\n",
      "38/38 - 9s - loss: 0.5461 - accuracy: 0.7306 - val_loss: 0.5181 - val_accuracy: 0.7143 - lr: 0.0050 - 9s/epoch - 236ms/step\n",
      "Epoch 36/100\n",
      "\n",
      "Epoch 00036: val_loss did not improve from 0.51810\n",
      "38/38 - 8s - loss: 0.5382 - accuracy: 0.7471 - val_loss: 0.5272 - val_accuracy: 0.7245 - lr: 0.0050 - 8s/epoch - 220ms/step\n",
      "Epoch 37/100\n",
      "\n",
      "Epoch 00037: val_loss did not improve from 0.51810\n",
      "38/38 - 8s - loss: 0.5358 - accuracy: 0.7603 - val_loss: 0.5226 - val_accuracy: 0.7143 - lr: 0.0050 - 8s/epoch - 223ms/step\n",
      "Epoch 38/100\n",
      "\n",
      "Epoch 00038: val_loss improved from 0.51810 to 0.51516, saving model to D:\\Documentos\\projeto convs\\models\\cropped\n",
      "38/38 - 9s - loss: 0.5383 - accuracy: 0.7339 - val_loss: 0.5152 - val_accuracy: 0.7143 - lr: 0.0050 - 9s/epoch - 237ms/step\n",
      "Epoch 39/100\n",
      "\n",
      "Epoch 00039: val_loss improved from 0.51516 to 0.51453, saving model to D:\\Documentos\\projeto convs\\models\\cropped\n",
      "38/38 - 9s - loss: 0.5419 - accuracy: 0.7140 - val_loss: 0.5145 - val_accuracy: 0.7041 - lr: 0.0050 - 9s/epoch - 238ms/step\n",
      "Epoch 40/100\n",
      "\n",
      "Epoch 00040: val_loss did not improve from 0.51453\n",
      "38/38 - 8s - loss: 0.5374 - accuracy: 0.7322 - val_loss: 0.5159 - val_accuracy: 0.7041 - lr: 0.0050 - 8s/epoch - 218ms/step\n",
      "Epoch 41/100\n",
      "\n",
      "Epoch 00041: val_loss did not improve from 0.51453\n",
      "38/38 - 8s - loss: 0.5341 - accuracy: 0.7273 - val_loss: 0.5169 - val_accuracy: 0.7143 - lr: 0.0050 - 8s/epoch - 220ms/step\n",
      "Epoch 42/100\n",
      "\n",
      "Epoch 00042: val_loss improved from 0.51453 to 0.51262, saving model to D:\\Documentos\\projeto convs\\models\\cropped\n",
      "38/38 - 9s - loss: 0.5365 - accuracy: 0.7388 - val_loss: 0.5126 - val_accuracy: 0.7143 - lr: 0.0050 - 9s/epoch - 238ms/step\n",
      "Epoch 43/100\n",
      "\n",
      "Epoch 00043: val_loss did not improve from 0.51262\n",
      "38/38 - 8s - loss: 0.5310 - accuracy: 0.7488 - val_loss: 0.5149 - val_accuracy: 0.7041 - lr: 0.0050 - 8s/epoch - 219ms/step\n",
      "Epoch 44/100\n",
      "\n",
      "Epoch 00044: val_loss improved from 0.51262 to 0.51149, saving model to D:\\Documentos\\projeto convs\\models\\cropped\n",
      "38/38 - 9s - loss: 0.5300 - accuracy: 0.7438 - val_loss: 0.5115 - val_accuracy: 0.7143 - lr: 0.0050 - 9s/epoch - 237ms/step\n",
      "Epoch 45/100\n",
      "\n",
      "Epoch 00045: val_loss improved from 0.51149 to 0.50991, saving model to D:\\Documentos\\projeto convs\\models\\cropped\n",
      "38/38 - 9s - loss: 0.5288 - accuracy: 0.7405 - val_loss: 0.5099 - val_accuracy: 0.7143 - lr: 0.0050 - 9s/epoch - 239ms/step\n",
      "Epoch 46/100\n",
      "\n",
      "Epoch 00046: val_loss improved from 0.50991 to 0.50969, saving model to D:\\Documentos\\projeto convs\\models\\cropped\n",
      "38/38 - 9s - loss: 0.5380 - accuracy: 0.7322 - val_loss: 0.5097 - val_accuracy: 0.7143 - lr: 0.0050 - 9s/epoch - 242ms/step\n",
      "Epoch 47/100\n",
      "\n",
      "Epoch 00047: val_loss improved from 0.50969 to 0.50839, saving model to D:\\Documentos\\projeto convs\\models\\cropped\n",
      "38/38 - 9s - loss: 0.5259 - accuracy: 0.7504 - val_loss: 0.5084 - val_accuracy: 0.7143 - lr: 0.0050 - 9s/epoch - 236ms/step\n",
      "Epoch 48/100\n",
      "\n",
      "Epoch 00048: val_loss did not improve from 0.50839\n",
      "38/38 - 8s - loss: 0.5270 - accuracy: 0.7537 - val_loss: 0.5107 - val_accuracy: 0.7041 - lr: 0.0050 - 8s/epoch - 220ms/step\n",
      "Epoch 49/100\n",
      "\n",
      "Epoch 00049: val_loss did not improve from 0.50839\n",
      "38/38 - 8s - loss: 0.5363 - accuracy: 0.7455 - val_loss: 0.5094 - val_accuracy: 0.7041 - lr: 0.0050 - 8s/epoch - 219ms/step\n",
      "Epoch 50/100\n",
      "\n",
      "Epoch 00050: val_loss improved from 0.50839 to 0.50785, saving model to D:\\Documentos\\projeto convs\\models\\cropped\n",
      "38/38 - 9s - loss: 0.5195 - accuracy: 0.7620 - val_loss: 0.5078 - val_accuracy: 0.7143 - lr: 0.0050 - 9s/epoch - 239ms/step\n",
      "Epoch 51/100\n",
      "\n",
      "Epoch 00051: val_loss did not improve from 0.50785\n",
      "38/38 - 8s - loss: 0.5229 - accuracy: 0.7438 - val_loss: 0.5085 - val_accuracy: 0.7041 - lr: 0.0050 - 8s/epoch - 221ms/step\n",
      "Epoch 52/100\n",
      "\n",
      "Epoch 00052: val_loss did not improve from 0.50785\n",
      "38/38 - 8s - loss: 0.5317 - accuracy: 0.7438 - val_loss: 0.5094 - val_accuracy: 0.7143 - lr: 0.0050 - 8s/epoch - 220ms/step\n",
      "Epoch 53/100\n",
      "\n",
      "Epoch 00053: val_loss improved from 0.50785 to 0.50193, saving model to D:\\Documentos\\projeto convs\\models\\cropped\n",
      "38/38 - 9s - loss: 0.5166 - accuracy: 0.7554 - val_loss: 0.5019 - val_accuracy: 0.7551 - lr: 0.0050 - 9s/epoch - 239ms/step\n",
      "Epoch 54/100\n",
      "\n",
      "Epoch 00054: val_loss did not improve from 0.50193\n",
      "38/38 - 8s - loss: 0.5247 - accuracy: 0.7256 - val_loss: 0.5160 - val_accuracy: 0.7143 - lr: 0.0050 - 8s/epoch - 219ms/step\n",
      "Epoch 55/100\n",
      "\n",
      "Epoch 00055: val_loss did not improve from 0.50193\n",
      "38/38 - 8s - loss: 0.5137 - accuracy: 0.7686 - val_loss: 0.5046 - val_accuracy: 0.7143 - lr: 0.0050 - 8s/epoch - 220ms/step\n",
      "Epoch 56/100\n",
      "\n",
      "Epoch 00056: val_loss improved from 0.50193 to 0.50001, saving model to D:\\Documentos\\projeto convs\\models\\cropped\n",
      "38/38 - 9s - loss: 0.5228 - accuracy: 0.7405 - val_loss: 0.5000 - val_accuracy: 0.7551 - lr: 0.0050 - 9s/epoch - 240ms/step\n",
      "Epoch 57/100\n",
      "\n",
      "Epoch 00057: val_loss did not improve from 0.50001\n",
      "38/38 - 8s - loss: 0.5150 - accuracy: 0.7587 - val_loss: 0.5011 - val_accuracy: 0.7245 - lr: 0.0050 - 8s/epoch - 219ms/step\n",
      "Epoch 58/100\n",
      "\n",
      "Epoch 00058: val_loss improved from 0.50001 to 0.49924, saving model to D:\\Documentos\\projeto convs\\models\\cropped\n",
      "38/38 - 9s - loss: 0.5296 - accuracy: 0.7405 - val_loss: 0.4992 - val_accuracy: 0.7449 - lr: 0.0050 - 9s/epoch - 238ms/step\n",
      "Epoch 59/100\n",
      "\n",
      "Epoch 00059: val_loss did not improve from 0.49924\n",
      "38/38 - 8s - loss: 0.5034 - accuracy: 0.7736 - val_loss: 0.5092 - val_accuracy: 0.7245 - lr: 0.0050 - 8s/epoch - 221ms/step\n",
      "Epoch 60/100\n",
      "\n",
      "Epoch 00060: val_loss improved from 0.49924 to 0.49740, saving model to D:\\Documentos\\projeto convs\\models\\cropped\n",
      "38/38 - 9s - loss: 0.5118 - accuracy: 0.7405 - val_loss: 0.4974 - val_accuracy: 0.7449 - lr: 0.0050 - 9s/epoch - 238ms/step\n",
      "Epoch 61/100\n",
      "\n",
      "Epoch 00061: val_loss did not improve from 0.49740\n",
      "38/38 - 8s - loss: 0.5191 - accuracy: 0.7488 - val_loss: 0.5085 - val_accuracy: 0.7245 - lr: 0.0050 - 8s/epoch - 220ms/step\n",
      "Epoch 62/100\n",
      "\n",
      "Epoch 00062: val_loss did not improve from 0.49740\n",
      "38/38 - 8s - loss: 0.5318 - accuracy: 0.7306 - val_loss: 0.5069 - val_accuracy: 0.7245 - lr: 0.0050 - 8s/epoch - 220ms/step\n",
      "Epoch 63/100\n",
      "\n",
      "Epoch 00063: val_loss did not improve from 0.49740\n",
      "\n",
      "Epoch 00063: ReduceLROnPlateau reducing learning rate to 0.0024999999441206455.\n",
      "38/38 - 8s - loss: 0.5275 - accuracy: 0.7405 - val_loss: 0.5014 - val_accuracy: 0.7143 - lr: 0.0050 - 8s/epoch - 220ms/step\n",
      "Epoch 64/100\n",
      "\n",
      "Epoch 00064: val_loss did not improve from 0.49740\n",
      "38/38 - 8s - loss: 0.5179 - accuracy: 0.7521 - val_loss: 0.5007 - val_accuracy: 0.7143 - lr: 0.0025 - 8s/epoch - 221ms/step\n",
      "Epoch 65/100\n",
      "\n",
      "Epoch 00065: val_loss did not improve from 0.49740\n",
      "38/38 - 8s - loss: 0.5107 - accuracy: 0.7537 - val_loss: 0.5057 - val_accuracy: 0.7245 - lr: 0.0025 - 8s/epoch - 219ms/step\n",
      "Epoch 66/100\n",
      "\n",
      "Epoch 00066: val_loss did not improve from 0.49740\n",
      "38/38 - 8s - loss: 0.5286 - accuracy: 0.7388 - val_loss: 0.5051 - val_accuracy: 0.7245 - lr: 0.0025 - 8s/epoch - 210ms/step\n",
      "Epoch 67/100\n",
      "\n",
      "Epoch 00067: val_loss did not improve from 0.49740\n",
      "\n",
      "Epoch 00067: ReduceLROnPlateau reducing learning rate to 0.0012499999720603228.\n",
      "38/38 - 8s - loss: 0.5136 - accuracy: 0.7587 - val_loss: 0.5047 - val_accuracy: 0.7245 - lr: 0.0025 - 8s/epoch - 205ms/step\n",
      "Epoch 68/100\n",
      "\n",
      "Epoch 00068: val_loss did not improve from 0.49740\n",
      "38/38 - 8s - loss: 0.5178 - accuracy: 0.7438 - val_loss: 0.5063 - val_accuracy: 0.7245 - lr: 0.0012 - 8s/epoch - 204ms/step\n",
      "Epoch 69/100\n",
      "\n",
      "Epoch 00069: val_loss did not improve from 0.49740\n",
      "38/38 - 8s - loss: 0.5134 - accuracy: 0.7438 - val_loss: 0.5063 - val_accuracy: 0.7245 - lr: 0.0012 - 8s/epoch - 204ms/step\n",
      "Epoch 70/100\n",
      "\n",
      "Epoch 00070: val_loss did not improve from 0.49740\n",
      "38/38 - 8s - loss: 0.5120 - accuracy: 0.7521 - val_loss: 0.5063 - val_accuracy: 0.7245 - lr: 0.0012 - 8s/epoch - 206ms/step\n",
      "Epoch 71/100\n",
      "\n",
      "Epoch 00071: val_loss did not improve from 0.49740\n",
      "\n",
      "Epoch 00071: ReduceLROnPlateau reducing learning rate to 0.0006249999860301614.\n",
      "38/38 - 8s - loss: 0.5101 - accuracy: 0.7802 - val_loss: 0.5058 - val_accuracy: 0.7245 - lr: 0.0012 - 8s/epoch - 206ms/step\n",
      "Epoch 72/100\n",
      "\n",
      "Epoch 00072: val_loss did not improve from 0.49740\n",
      "38/38 - 8s - loss: 0.5122 - accuracy: 0.7636 - val_loss: 0.5053 - val_accuracy: 0.7245 - lr: 6.2500e-04 - 8s/epoch - 205ms/step\n",
      "Epoch 73/100\n",
      "\n",
      "Epoch 00073: val_loss did not improve from 0.49740\n",
      "38/38 - 8s - loss: 0.5142 - accuracy: 0.7521 - val_loss: 0.5057 - val_accuracy: 0.7245 - lr: 6.2500e-04 - 8s/epoch - 204ms/step\n",
      "Epoch 74/100\n",
      "\n",
      "Epoch 00074: val_loss did not improve from 0.49740\n",
      "38/38 - 8s - loss: 0.5171 - accuracy: 0.7421 - val_loss: 0.5051 - val_accuracy: 0.7245 - lr: 6.2500e-04 - 8s/epoch - 205ms/step\n",
      "Epoch 75/100\n",
      "\n",
      "Epoch 00075: val_loss did not improve from 0.49740\n",
      "\n",
      "Epoch 00075: ReduceLROnPlateau reducing learning rate to 0.0003124999930150807.\n",
      "38/38 - 8s - loss: 0.5086 - accuracy: 0.7603 - val_loss: 0.5042 - val_accuracy: 0.7245 - lr: 6.2500e-04 - 8s/epoch - 205ms/step\n"
     ]
    }
   ],
   "source": [
    "lr = kt.HyperParameters()\n",
    "lr.Fixed(name = 'learning rate', value = 1e-3)\n",
    "model = get_model(lr)\n",
    "history = model.fit(full_train_ds, batch_size = batch_size, epochs = 100, verbose = 2, validation_data = test_ds, callbacks = callbacks)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 576x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "model.load_weights(best_model_path)\n",
    "df = pd.DataFrame(history.history)[['loss', 'val_loss']]\n",
    "pd.DataFrame(df).plot(figsize=(8,5))\n",
    "plt.xlabel('Epochs')\n",
    "plt.savefig('fullcroppedloss.png')\n",
    "plt.show()\n",
    "\n",
    "#adicionar matriz de confusão\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 576x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "df = pd.DataFrame(history.history)[['accuracy', 'val_accuracy']]\n",
    "pd.DataFrame(df).plot(figsize=(8,5))\n",
    "plt.xlabel('Epochs')\n",
    "plt.savefig('fullcroppedaccuracy.png')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.plotly.v1+json": {
       "config": {
        "plotlyServerURL": "https://plot.ly"
       },
       "data": [
        {
         "hovertemplate": "False Positive Rate=%{x}<br>True Positive Rate=%{y}<extra></extra>",
         "legendgroup": "",
         "line": {
          "color": "#636efa"
         },
         "marker": {
          "symbol": "circle"
         },
         "mode": "lines",
         "name": "",
         "orientation": "v",
         "showlegend": false,
         "stackgroup": "1",
         "type": "scatter",
         "x": [
          0,
          0,
          0,
          0.022222222222222223,
          0.022222222222222223,
          0.044444444444444446,
          0.044444444444444446,
          0.06666666666666667,
          0.06666666666666667,
          0.08888888888888889,
          0.08888888888888889,
          0.1111111111111111,
          0.1111111111111111,
          0.13333333333333333,
          0.13333333333333333,
          0.17777777777777778,
          0.17777777777777778,
          0.2222222222222222,
          0.2222222222222222,
          0.28888888888888886,
          0.28888888888888886,
          0.3333333333333333,
          0.3333333333333333,
          0.35555555555555557,
          0.35555555555555557,
          0.5333333333333333,
          0.5333333333333333,
          0.6444444444444445,
          0.6444444444444445,
          0.8222222222222222,
          0.8222222222222222,
          1
         ],
         "xaxis": "x",
         "y": [
          0,
          0.018867924528301886,
          0.32075471698113206,
          0.32075471698113206,
          0.37735849056603776,
          0.37735849056603776,
          0.41509433962264153,
          0.41509433962264153,
          0.4528301886792453,
          0.4528301886792453,
          0.4716981132075472,
          0.4716981132075472,
          0.5283018867924528,
          0.5283018867924528,
          0.5849056603773585,
          0.5849056603773585,
          0.660377358490566,
          0.660377358490566,
          0.8301886792452831,
          0.8301886792452831,
          0.8867924528301887,
          0.8867924528301887,
          0.9056603773584906,
          0.9056603773584906,
          0.9433962264150944,
          0.9433962264150944,
          0.9622641509433962,
          0.9622641509433962,
          0.9811320754716981,
          0.9811320754716981,
          1,
          1
         ],
         "yaxis": "y"
        }
       ],
       "layout": {
        "height": 500,
        "legend": {
         "tracegroupgap": 0
        },
        "shapes": [
         {
          "line": {
           "dash": "dash"
          },
          "type": "line",
          "x0": 0,
          "x1": 1,
          "y0": 0,
          "y1": 1
         }
        ],
        "template": {
         "data": {
          "bar": [
           {
            "error_x": {
             "color": "#2a3f5f"
            },
            "error_y": {
             "color": "#2a3f5f"
            },
            "marker": {
             "line": {
              "color": "#E5ECF6",
              "width": 0.5
             },
             "pattern": {
              "fillmode": "overlay",
              "size": 10,
              "solidity": 0.2
             }
            },
            "type": "bar"
           }
          ],
          "barpolar": [
           {
            "marker": {
             "line": {
              "color": "#E5ECF6",
              "width": 0.5
             },
             "pattern": {
              "fillmode": "overlay",
              "size": 10,
              "solidity": 0.2
             }
            },
            "type": "barpolar"
           }
          ],
          "carpet": [
           {
            "aaxis": {
             "endlinecolor": "#2a3f5f",
             "gridcolor": "white",
             "linecolor": "white",
             "minorgridcolor": "white",
             "startlinecolor": "#2a3f5f"
            },
            "baxis": {
             "endlinecolor": "#2a3f5f",
             "gridcolor": "white",
             "linecolor": "white",
             "minorgridcolor": "white",
             "startlinecolor": "#2a3f5f"
            },
            "type": "carpet"
           }
          ],
          "choropleth": [
           {
            "colorbar": {
             "outlinewidth": 0,
             "ticks": ""
            },
            "type": "choropleth"
           }
          ],
          "contour": [
           {
            "colorbar": {
             "outlinewidth": 0,
             "ticks": ""
            },
            "colorscale": [
             [
              0,
              "#0d0887"
             ],
             [
              0.1111111111111111,
              "#46039f"
             ],
             [
              0.2222222222222222,
              "#7201a8"
             ],
             [
              0.3333333333333333,
              "#9c179e"
             ],
             [
              0.4444444444444444,
              "#bd3786"
             ],
             [
              0.5555555555555556,
              "#d8576b"
             ],
             [
              0.6666666666666666,
              "#ed7953"
             ],
             [
              0.7777777777777778,
              "#fb9f3a"
             ],
             [
              0.8888888888888888,
              "#fdca26"
             ],
             [
              1,
              "#f0f921"
             ]
            ],
            "type": "contour"
           }
          ],
          "contourcarpet": [
           {
            "colorbar": {
             "outlinewidth": 0,
             "ticks": ""
            },
            "type": "contourcarpet"
           }
          ],
          "heatmap": [
           {
            "colorbar": {
             "outlinewidth": 0,
             "ticks": ""
            },
            "colorscale": [
             [
              0,
              "#0d0887"
             ],
             [
              0.1111111111111111,
              "#46039f"
             ],
             [
              0.2222222222222222,
              "#7201a8"
             ],
             [
              0.3333333333333333,
              "#9c179e"
             ],
             [
              0.4444444444444444,
              "#bd3786"
             ],
             [
              0.5555555555555556,
              "#d8576b"
             ],
             [
              0.6666666666666666,
              "#ed7953"
             ],
             [
              0.7777777777777778,
              "#fb9f3a"
             ],
             [
              0.8888888888888888,
              "#fdca26"
             ],
             [
              1,
              "#f0f921"
             ]
            ],
            "type": "heatmap"
           }
          ],
          "heatmapgl": [
           {
            "colorbar": {
             "outlinewidth": 0,
             "ticks": ""
            },
            "colorscale": [
             [
              0,
              "#0d0887"
             ],
             [
              0.1111111111111111,
              "#46039f"
             ],
             [
              0.2222222222222222,
              "#7201a8"
             ],
             [
              0.3333333333333333,
              "#9c179e"
             ],
             [
              0.4444444444444444,
              "#bd3786"
             ],
             [
              0.5555555555555556,
              "#d8576b"
             ],
             [
              0.6666666666666666,
              "#ed7953"
             ],
             [
              0.7777777777777778,
              "#fb9f3a"
             ],
             [
              0.8888888888888888,
              "#fdca26"
             ],
             [
              1,
              "#f0f921"
             ]
            ],
            "type": "heatmapgl"
           }
          ],
          "histogram": [
           {
            "marker": {
             "pattern": {
              "fillmode": "overlay",
              "size": 10,
              "solidity": 0.2
             }
            },
            "type": "histogram"
           }
          ],
          "histogram2d": [
           {
            "colorbar": {
             "outlinewidth": 0,
             "ticks": ""
            },
            "colorscale": [
             [
              0,
              "#0d0887"
             ],
             [
              0.1111111111111111,
              "#46039f"
             ],
             [
              0.2222222222222222,
              "#7201a8"
             ],
             [
              0.3333333333333333,
              "#9c179e"
             ],
             [
              0.4444444444444444,
              "#bd3786"
             ],
             [
              0.5555555555555556,
              "#d8576b"
             ],
             [
              0.6666666666666666,
              "#ed7953"
             ],
             [
              0.7777777777777778,
              "#fb9f3a"
             ],
             [
              0.8888888888888888,
              "#fdca26"
             ],
             [
              1,
              "#f0f921"
             ]
            ],
            "type": "histogram2d"
           }
          ],
          "histogram2dcontour": [
           {
            "colorbar": {
             "outlinewidth": 0,
             "ticks": ""
            },
            "colorscale": [
             [
              0,
              "#0d0887"
             ],
             [
              0.1111111111111111,
              "#46039f"
             ],
             [
              0.2222222222222222,
              "#7201a8"
             ],
             [
              0.3333333333333333,
              "#9c179e"
             ],
             [
              0.4444444444444444,
              "#bd3786"
             ],
             [
              0.5555555555555556,
              "#d8576b"
             ],
             [
              0.6666666666666666,
              "#ed7953"
             ],
             [
              0.7777777777777778,
              "#fb9f3a"
             ],
             [
              0.8888888888888888,
              "#fdca26"
             ],
             [
              1,
              "#f0f921"
             ]
            ],
            "type": "histogram2dcontour"
           }
          ],
          "mesh3d": [
           {
            "colorbar": {
             "outlinewidth": 0,
             "ticks": ""
            },
            "type": "mesh3d"
           }
          ],
          "parcoords": [
           {
            "line": {
             "colorbar": {
              "outlinewidth": 0,
              "ticks": ""
             }
            },
            "type": "parcoords"
           }
          ],
          "pie": [
           {
            "automargin": true,
            "type": "pie"
           }
          ],
          "scatter": [
           {
            "marker": {
             "colorbar": {
              "outlinewidth": 0,
              "ticks": ""
             }
            },
            "type": "scatter"
           }
          ],
          "scatter3d": [
           {
            "line": {
             "colorbar": {
              "outlinewidth": 0,
              "ticks": ""
             }
            },
            "marker": {
             "colorbar": {
              "outlinewidth": 0,
              "ticks": ""
             }
            },
            "type": "scatter3d"
           }
          ],
          "scattercarpet": [
           {
            "marker": {
             "colorbar": {
              "outlinewidth": 0,
              "ticks": ""
             }
            },
            "type": "scattercarpet"
           }
          ],
          "scattergeo": [
           {
            "marker": {
             "colorbar": {
              "outlinewidth": 0,
              "ticks": ""
             }
            },
            "type": "scattergeo"
           }
          ],
          "scattergl": [
           {
            "marker": {
             "colorbar": {
              "outlinewidth": 0,
              "ticks": ""
             }
            },
            "type": "scattergl"
           }
          ],
          "scattermapbox": [
           {
            "marker": {
             "colorbar": {
              "outlinewidth": 0,
              "ticks": ""
             }
            },
            "type": "scattermapbox"
           }
          ],
          "scatterpolar": [
           {
            "marker": {
             "colorbar": {
              "outlinewidth": 0,
              "ticks": ""
             }
            },
            "type": "scatterpolar"
           }
          ],
          "scatterpolargl": [
           {
            "marker": {
             "colorbar": {
              "outlinewidth": 0,
              "ticks": ""
             }
            },
            "type": "scatterpolargl"
           }
          ],
          "scatterternary": [
           {
            "marker": {
             "colorbar": {
              "outlinewidth": 0,
              "ticks": ""
             }
            },
            "type": "scatterternary"
           }
          ],
          "surface": [
           {
            "colorbar": {
             "outlinewidth": 0,
             "ticks": ""
            },
            "colorscale": [
             [
              0,
              "#0d0887"
             ],
             [
              0.1111111111111111,
              "#46039f"
             ],
             [
              0.2222222222222222,
              "#7201a8"
             ],
             [
              0.3333333333333333,
              "#9c179e"
             ],
             [
              0.4444444444444444,
              "#bd3786"
             ],
             [
              0.5555555555555556,
              "#d8576b"
             ],
             [
              0.6666666666666666,
              "#ed7953"
             ],
             [
              0.7777777777777778,
              "#fb9f3a"
             ],
             [
              0.8888888888888888,
              "#fdca26"
             ],
             [
              1,
              "#f0f921"
             ]
            ],
            "type": "surface"
           }
          ],
          "table": [
           {
            "cells": {
             "fill": {
              "color": "#EBF0F8"
             },
             "line": {
              "color": "white"
             }
            },
            "header": {
             "fill": {
              "color": "#C8D4E3"
             },
             "line": {
              "color": "white"
             }
            },
            "type": "table"
           }
          ]
         },
         "layout": {
          "annotationdefaults": {
           "arrowcolor": "#2a3f5f",
           "arrowhead": 0,
           "arrowwidth": 1
          },
          "autotypenumbers": "strict",
          "coloraxis": {
           "colorbar": {
            "outlinewidth": 0,
            "ticks": ""
           }
          },
          "colorscale": {
           "diverging": [
            [
             0,
             "#8e0152"
            ],
            [
             0.1,
             "#c51b7d"
            ],
            [
             0.2,
             "#de77ae"
            ],
            [
             0.3,
             "#f1b6da"
            ],
            [
             0.4,
             "#fde0ef"
            ],
            [
             0.5,
             "#f7f7f7"
            ],
            [
             0.6,
             "#e6f5d0"
            ],
            [
             0.7,
             "#b8e186"
            ],
            [
             0.8,
             "#7fbc41"
            ],
            [
             0.9,
             "#4d9221"
            ],
            [
             1,
             "#276419"
            ]
           ],
           "sequential": [
            [
             0,
             "#0d0887"
            ],
            [
             0.1111111111111111,
             "#46039f"
            ],
            [
             0.2222222222222222,
             "#7201a8"
            ],
            [
             0.3333333333333333,
             "#9c179e"
            ],
            [
             0.4444444444444444,
             "#bd3786"
            ],
            [
             0.5555555555555556,
             "#d8576b"
            ],
            [
             0.6666666666666666,
             "#ed7953"
            ],
            [
             0.7777777777777778,
             "#fb9f3a"
            ],
            [
             0.8888888888888888,
             "#fdca26"
            ],
            [
             1,
             "#f0f921"
            ]
           ],
           "sequentialminus": [
            [
             0,
             "#0d0887"
            ],
            [
             0.1111111111111111,
             "#46039f"
            ],
            [
             0.2222222222222222,
             "#7201a8"
            ],
            [
             0.3333333333333333,
             "#9c179e"
            ],
            [
             0.4444444444444444,
             "#bd3786"
            ],
            [
             0.5555555555555556,
             "#d8576b"
            ],
            [
             0.6666666666666666,
             "#ed7953"
            ],
            [
             0.7777777777777778,
             "#fb9f3a"
            ],
            [
             0.8888888888888888,
             "#fdca26"
            ],
            [
             1,
             "#f0f921"
            ]
           ]
          },
          "colorway": [
           "#636efa",
           "#EF553B",
           "#00cc96",
           "#ab63fa",
           "#FFA15A",
           "#19d3f3",
           "#FF6692",
           "#B6E880",
           "#FF97FF",
           "#FECB52"
          ],
          "font": {
           "color": "#2a3f5f"
          },
          "geo": {
           "bgcolor": "white",
           "lakecolor": "white",
           "landcolor": "#E5ECF6",
           "showlakes": true,
           "showland": true,
           "subunitcolor": "white"
          },
          "hoverlabel": {
           "align": "left"
          },
          "hovermode": "closest",
          "mapbox": {
           "style": "light"
          },
          "paper_bgcolor": "white",
          "plot_bgcolor": "#E5ECF6",
          "polar": {
           "angularaxis": {
            "gridcolor": "white",
            "linecolor": "white",
            "ticks": ""
           },
           "bgcolor": "#E5ECF6",
           "radialaxis": {
            "gridcolor": "white",
            "linecolor": "white",
            "ticks": ""
           }
          },
          "scene": {
           "xaxis": {
            "backgroundcolor": "#E5ECF6",
            "gridcolor": "white",
            "gridwidth": 2,
            "linecolor": "white",
            "showbackground": true,
            "ticks": "",
            "zerolinecolor": "white"
           },
           "yaxis": {
            "backgroundcolor": "#E5ECF6",
            "gridcolor": "white",
            "gridwidth": 2,
            "linecolor": "white",
            "showbackground": true,
            "ticks": "",
            "zerolinecolor": "white"
           },
           "zaxis": {
            "backgroundcolor": "#E5ECF6",
            "gridcolor": "white",
            "gridwidth": 2,
            "linecolor": "white",
            "showbackground": true,
            "ticks": "",
            "zerolinecolor": "white"
           }
          },
          "shapedefaults": {
           "line": {
            "color": "#2a3f5f"
           }
          },
          "ternary": {
           "aaxis": {
            "gridcolor": "white",
            "linecolor": "white",
            "ticks": ""
           },
           "baxis": {
            "gridcolor": "white",
            "linecolor": "white",
            "ticks": ""
           },
           "bgcolor": "#E5ECF6",
           "caxis": {
            "gridcolor": "white",
            "linecolor": "white",
            "ticks": ""
           }
          },
          "title": {
           "x": 0.05
          },
          "xaxis": {
           "automargin": true,
           "gridcolor": "white",
           "linecolor": "white",
           "ticks": "",
           "title": {
            "standoff": 15
           },
           "zerolinecolor": "white",
           "zerolinewidth": 2
          },
          "yaxis": {
           "automargin": true,
           "gridcolor": "white",
           "linecolor": "white",
           "ticks": "",
           "title": {
            "standoff": 15
           },
           "zerolinecolor": "white",
           "zerolinewidth": 2
          }
         }
        },
        "title": {
         "text": "ROC Curve (AUC=0.8541)"
        },
        "width": 700,
        "xaxis": {
         "anchor": "y",
         "constrain": "domain",
         "domain": [
          0,
          1
         ],
         "title": {
          "text": "False Positive Rate"
         }
        },
        "yaxis": {
         "anchor": "x",
         "domain": [
          0,
          1
         ],
         "scaleanchor": "x",
         "scaleratio": 1,
         "title": {
          "text": "True Positive Rate"
         }
        }
       }
      }
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "def get_binary_labels(ds, model):\n",
    "    ''' TF.dataset --> numpy.array\n",
    "    takes tensorflow dataset for data with binary classification and returns array\n",
    "    with an int representing the class for each element'''\n",
    "    labels = np.array([])\n",
    "    preds = np.array([])\n",
    "    for x, y in ds.as_numpy_iterator():\n",
    "        preds = np.append(preds, model.predict(x))\n",
    "        labels = np.append(labels, y)\n",
    "    labels = np.array([int(label) for label in labels])\n",
    "    return preds, labels\n",
    "\n",
    "preds, y = get_binary_labels(test_ds, model)\n",
    "\n",
    "import matplotlib.pyplot as plt\n",
    "from sklearn.metrics import confusion_matrix, classification_report, auc,  roc_curve\n",
    "import plotly.express as px\n",
    "\n",
    "\n",
    "fpr, tpr, thresholds = roc_curve(y, preds)\n",
    "fig = px.area(\n",
    "    x=fpr, y=tpr,\n",
    "    title=f'ROC Curve (AUC={auc(fpr, tpr):.4f})',\n",
    "    labels=dict(x='False Positive Rate', y='True Positive Rate'),\n",
    "    width=700, height=500\n",
    ")\n",
    "fig.add_shape(\n",
    "    type='line', line=dict(dash='dash'),\n",
    "    x0=0, x1=1, y0=0, y1=1\n",
    ")\n",
    "\n",
    "fig.update_yaxes(scaleanchor=\"x\", scaleratio=1)\n",
    "fig.update_xaxes(constrain='domain')\n",
    "fig.show()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "<Figure size 720x504 with 0 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.80      0.78      0.79        45\n",
      "           1       0.81      0.83      0.82        53\n",
      "\n",
      "    accuracy                           0.81        98\n",
      "   macro avg       0.81      0.80      0.80        98\n",
      "weighted avg       0.81      0.81      0.81        98\n",
      "\n"
     ]
    }
   ],
   "source": [
    "import seaborn as sns\n",
    "\n",
    "dists = np.array([(np.sqrt((0 - x)**2 + (1-y)**2)) for x, y in zip(fpr, tpr)])\n",
    "cutoff_idx = dists.argmin()\n",
    "cutoff = thresholds[cutoff_idx]\n",
    "\n",
    "#Selecionando cutoff\n",
    "\n",
    "preds2 = np.array([1 if x >= cutoff else 0 for x in preds])\n",
    "cm = confusion_matrix(y, preds2)\n",
    "#turn this into a dataframe\n",
    "matrix_df = pd.DataFrame(cm)\n",
    "#plot the result\n",
    "ax = plt.axes()\n",
    "sns.set(font_scale=1.3)\n",
    "plt.figure(figsize=(10,7))\n",
    "sns.heatmap(matrix_df, annot=True, fmt=\"g\", ax=ax, cmap=\"magma\")\n",
    "#set axis titles\n",
    "ax.set_title('Confusion Matrix - Dense121')\n",
    "ax.set_xlabel(\"Predicted label\", fontsize =15)\n",
    "ax.set_ylabel(\"True Label\", fontsize=15)\n",
    "plt.savefig('fullcroppedcm.png')\n",
    "\n",
    "plt.show()\n",
    "print(classification_report(y, preds2))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\\begin{tabular}{lrrrr}\n",
      "\\toprule\n",
      "{} &  precision &    recall &  f1-score &    support \\\\\n",
      "\\midrule\n",
      "0            &   0.795455 &  0.777778 &  0.786517 &  45.000000 \\\\\n",
      "1            &   0.814815 &  0.830189 &  0.822430 &  53.000000 \\\\\n",
      "accuracy     &   0.806122 &  0.806122 &  0.806122 &   0.806122 \\\\\n",
      "macro avg    &   0.805135 &  0.803983 &  0.804473 &  98.000000 \\\\\n",
      "weighted avg &   0.805925 &  0.806122 &  0.805939 &  98.000000 \\\\\n",
      "\\bottomrule\n",
      "\\end{tabular}\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(pd.DataFrame(classification_report(y, preds2, output_dict=True)).transpose().to_latex())"
   ]
  }
 ],
 "metadata": {
  "interpreter": {
   "hash": "81501fb6968db567c48e2f6497df9bfcfab897da0876ab26c596c8e980f4fbda"
  },
  "kernelspec": {
   "display_name": "Python 3.9.4 64-bit",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.9.4"
  },
  "orig_nbformat": 4
 },
 "nbformat": 4,
 "nbformat_minor": 2
}