ex-5-custom-convnets-computer-vision.ipynb

{"cells":[{"source":"<a href=\"https://www.kaggle.com/code/parisanahmadi/ex-5-custom-convnets-computer-vision?scriptVersionId=103353603\" target=\"_blank\"><img align=\"left\" alt=\"Kaggle\" title=\"Open in Kaggle\" src=\"https://kaggle.com/static/images/open-in-kaggle.svg\"></a>","metadata":{},"cell_type":"markdown","outputs":[],"execution_count":0},{"cell_type":"markdown","id":"2404c99c","metadata":{"papermill":{"duration":0.004654,"end_time":"2022-08-15T14:12:25.585117","exception":false,"start_time":"2022-08-15T14:12:25.580463","status":"completed"},"tags":[]},"source":["![](https://ammarjaved.com/storage/posts/what-is-computer-vision-applications-of-computer-vision.png)"]},{"cell_type":"markdown","id":"5c7b5eb3","metadata":{"papermill":{"duration":0.003558,"end_time":"2022-08-15T14:12:25.592567","exception":false,"start_time":"2022-08-15T14:12:25.589009","status":"completed"},"tags":[]},"source":["# **This notebook is an exercise in the [Computer Vision](https://www.kaggle.com/learn/computer-vision) course.  You can reference the tutorial at [this link](https://www.kaggle.com/ryanholbrook/custom-convnets).**\n","\n","---\n"]},{"cell_type":"markdown","id":"d45d536c","metadata":{"papermill":{"duration":0.003379,"end_time":"2022-08-15T14:12:25.599517","exception":false,"start_time":"2022-08-15T14:12:25.596138","status":"completed"},"tags":[]},"source":["# Introduction #\n","\n","In these exercises, you'll build a custom convnet with performance competitive to the VGG16 model from Lesson 1.\n","\n","Get started by running the code cell below."]},{"cell_type":"code","execution_count":1,"id":"836b73f6","metadata":{"execution":{"iopub.execute_input":"2022-08-15T14:12:25.609434Z","iopub.status.busy":"2022-08-15T14:12:25.608892Z","iopub.status.idle":"2022-08-15T14:12:39.992869Z","shell.execute_reply":"2022-08-15T14:12:39.99192Z"},"papermill":{"duration":14.392061,"end_time":"2022-08-15T14:12:39.995165","exception":false,"start_time":"2022-08-15T14:12:25.603104","status":"completed"},"tags":[]},"outputs":[{"name":"stdout","output_type":"stream","text":["Found 5117 files belonging to 2 classes.\n"]},{"name":"stderr","output_type":"stream","text":["2022-08-15 14:12:33.996973: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:937] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero\n","2022-08-15 14:12:34.117641: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:937] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero\n","2022-08-15 14:12:34.118401: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:937] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero\n","2022-08-15 14:12:34.131036: I tensorflow/core/platform/cpu_feature_guard.cc:142] This TensorFlow binary is optimized with oneAPI Deep Neural Network Library (oneDNN) to use the following CPU instructions in performance-critical operations:  AVX2 AVX512F FMA\n","To enable them in other operations, rebuild TensorFlow with the appropriate compiler flags.\n","2022-08-15 14:12:34.131358: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:937] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero\n","2022-08-15 14:12:34.132450: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:937] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero\n","2022-08-15 14:12:34.133417: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:937] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero\n","2022-08-15 14:12:36.415450: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:937] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero\n","2022-08-15 14:12:36.416307: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:937] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero\n","2022-08-15 14:12:36.417048: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:937] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero\n","2022-08-15 14:12:36.417636: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1510] Created device /job:localhost/replica:0/task:0/device:GPU:0 with 15403 MB memory:  -> device: 0, name: Tesla P100-PCIE-16GB, pci bus id: 0000:00:04.0, compute capability: 6.0\n"]},{"name":"stdout","output_type":"stream","text":["Found 5051 files belonging to 2 classes.\n"]}],"source":["# Setup feedback system\n","from learntools.core import binder\n","binder.bind(globals())\n","from learntools.computer_vision.ex5 import *\n","\n","# Imports\n","import os, warnings\n","import matplotlib.pyplot as plt\n","from matplotlib import gridspec\n","\n","import numpy as np\n","import tensorflow as tf\n","from tensorflow.keras.preprocessing import image_dataset_from_directory\n","\n","# Reproducability\n","def set_seed(seed=31415):\n","    np.random.seed(seed)\n","    tf.random.set_seed(seed)\n","    os.environ['PYTHONHASHSE/ED'] = str(seed)\n","    os.environ['TF_DETERMINISTIC_OPS'] = '1'\n","set_seed()\n","\n","# Set Matplotlib defaults\n","plt.rc('figure', autolayout=True)\n","plt.rc('axes', labelweight='bold', labelsize='large',\n","       titleweight='bold', titlesize=18, titlepad=10)\n","plt.rc('image', cmap='magma')\n","warnings.filterwarnings(\"ignore\") # to clean up output cells\n","\n","\n","# Load training and validation sets\n","ds_train_ = image_dataset_from_directory(\n","    '../input/car-or-truck/train',\n","    labels='inferred',\n","    label_mode='binary',\n","    image_size=[128, 128],\n","    interpolation='nearest',\n","    batch_size=64,\n","    shuffle=True,\n",")\n","ds_valid_ = image_dataset_from_directory(\n","    '../input/car-or-truck/valid',\n","    labels='inferred',\n","    label_mode='binary',\n","    image_size=[128, 128],\n","    interpolation='nearest',\n","    batch_size=64,\n","    shuffle=False,\n",")\n","\n","# Data Pipeline\n","def convert_to_float(image, label):\n","    image = tf.image.convert_image_dtype(image, dtype=tf.float32)\n","    return image, label\n","\n","AUTOTUNE = tf.data.experimental.AUTOTUNE\n","ds_train = (\n","    ds_train_\n","    .map(convert_to_float)\n","    .cache()\n","    .prefetch(buffer_size=AUTOTUNE)\n",")\n","ds_valid = (\n","    ds_valid_\n","    .map(convert_to_float)\n","    .cache()\n","    .prefetch(buffer_size=AUTOTUNE)\n",")\n"]},{"cell_type":"markdown","id":"b91cf44c","metadata":{"papermill":{"duration":0.003828,"end_time":"2022-08-15T14:12:40.003123","exception":false,"start_time":"2022-08-15T14:12:39.999295","status":"completed"},"tags":[]},"source":["# Design a Convnet #\n","\n","Let's design a convolutional network with a block architecture like we saw in the tutorial. The model from the example had three blocks, each with a single convolutional layer. Its performance on the \"Car or Truck\" problem was okay, but far from what the pretrained VGG16 could achieve. It might be that our simple network lacks the ability to extract sufficiently complex features. We could try improving the model either by adding more blocks or by adding convolutions to the blocks we have.\n","\n","Let's go with the second approach. We'll keep the three block structure, but increase the number of `Conv2D` layer in the second block to two, and in the third block to three.\n","\n","<figure>\n","<!-- <img src=\"./images/2-convmodel-2.png\" width=\"250\" alt=\"Diagram of a convolutional model.\"> -->\n","<img src=\"https://i.imgur.com/Vko6nCK.png\" width=\"250\" alt=\"Diagram of a convolutional model.\">\n","</figure>\n","\n","# 1) Define Model #\n","\n","Given the diagram above, complete the model by defining the layers of the third block."]},{"cell_type":"code","execution_count":2,"id":"5d957e38","metadata":{"execution":{"iopub.execute_input":"2022-08-15T14:12:40.012594Z","iopub.status.busy":"2022-08-15T14:12:40.012266Z","iopub.status.idle":"2022-08-15T14:12:40.119329Z","shell.execute_reply":"2022-08-15T14:12:40.118382Z"},"lines_to_next_cell":0,"papermill":{"duration":0.113852,"end_time":"2022-08-15T14:12:40.121308","exception":false,"start_time":"2022-08-15T14:12:40.007456","status":"completed"},"tags":[]},"outputs":[{"data":{"application/javascript":["parent.postMessage({\"jupyterEvent\": \"custom.exercise_interaction\", \"data\": {\"outcomeType\": 1, \"valueTowardsCompletion\": 0.3333333333333333, \"interactionType\": 1, \"questionType\": 2, \"questionId\": \"1_Q1\", \"learnToolsVersion\": \"0.3.4\", \"failureMessage\": \"\", \"exceptionClass\": \"\", \"trace\": \"\"}}, \"*\")"],"text/plain":["<IPython.core.display.Javascript object>"]},"metadata":{},"output_type":"display_data"},{"data":{"text/markdown":["<span style=\"color:#33cc33\">Correct</span>"],"text/plain":["Correct"]},"metadata":{},"output_type":"display_data"}],"source":["from tensorflow import keras\n","from tensorflow.keras import layers\n","\n","model = keras.Sequential([\n","    # Block One\n","    layers.Conv2D(filters=32, kernel_size=3, activation='relu', padding='same',\n","                  input_shape=[128, 128, 3]),\n","    layers.MaxPool2D(),\n","\n","    # Block Two\n","    layers.Conv2D(filters=64, kernel_size=3, activation='relu', padding='same'),\n","    layers.MaxPool2D(),\n","\n","    # Block Three\n","    # YOUR CODE HERE\n","    layers.Conv2D(filters=128, kernel_size=3, activation=\"relu\", padding= \"same\"),\n","    layers.Conv2D(filters=128, kernel_size=3, activation=\"relu\", padding= \"same\"),\n","    layers.MaxPool2D(),\n","    \n","\n","    # Head\n","    layers.Flatten(),\n","    layers.Dense(6, activation='relu'),\n","    layers.Dropout(0.2),\n","    layers.Dense(1, activation='sigmoid'),\n","])\n","\n","# Check your answer\n","q_1.check()"]},{"cell_type":"code","execution_count":3,"id":"1797fc43","metadata":{"execution":{"iopub.execute_input":"2022-08-15T14:12:40.131821Z","iopub.status.busy":"2022-08-15T14:12:40.130967Z","iopub.status.idle":"2022-08-15T14:12:40.135166Z","shell.execute_reply":"2022-08-15T14:12:40.134332Z"},"papermill":{"duration":0.011211,"end_time":"2022-08-15T14:12:40.137199","exception":false,"start_time":"2022-08-15T14:12:40.125988","status":"completed"},"tags":[]},"outputs":[],"source":["# Lines below will give you a hint or solution code\n","#q_1.hint()\n","#q_1.solution()"]},{"cell_type":"markdown","id":"de5cdd80","metadata":{"papermill":{"duration":0.003857,"end_time":"2022-08-15T14:12:40.145024","exception":false,"start_time":"2022-08-15T14:12:40.141167","status":"completed"},"tags":[]},"source":["# 2) Compile #\n","\n","To prepare for training, compile the model with an appropriate loss and accuracy metric for the \"Car or Truck\" dataset."]},{"cell_type":"code","execution_count":4,"id":"b21d7e7b","metadata":{"execution":{"iopub.execute_input":"2022-08-15T14:12:40.154258Z","iopub.status.busy":"2022-08-15T14:12:40.153994Z","iopub.status.idle":"2022-08-15T14:12:40.1682Z","shell.execute_reply":"2022-08-15T14:12:40.167225Z"},"lines_to_next_cell":0,"papermill":{"duration":0.021147,"end_time":"2022-08-15T14:12:40.170145","exception":false,"start_time":"2022-08-15T14:12:40.148998","status":"completed"},"tags":[]},"outputs":[{"data":{"application/javascript":["parent.postMessage({\"jupyterEvent\": \"custom.exercise_interaction\", \"data\": {\"outcomeType\": 1, \"valueTowardsCompletion\": 0.3333333333333333, \"interactionType\": 1, \"questionType\": 2, \"questionId\": \"2_Q2\", \"learnToolsVersion\": \"0.3.4\", \"failureMessage\": \"\", \"exceptionClass\": \"\", \"trace\": \"\"}}, \"*\")"],"text/plain":["<IPython.core.display.Javascript object>"]},"metadata":{},"output_type":"display_data"},{"data":{"text/markdown":["<span style=\"color:#33cc33\">Correct</span>"],"text/plain":["Correct"]},"metadata":{},"output_type":"display_data"}],"source":["model.compile(\n","    optimizer=tf.keras.optimizers.Adam(epsilon=0.01),\n","    # YOUR CODE HERE: Add loss and metric\n","    loss = \"binary_crossentropy\",\n","    metrics = [\"binary_accuracy\"],\n",")\n","\n","# Check your answer\n","q_2.check()"]},{"cell_type":"code","execution_count":5,"id":"06b43bab","metadata":{"execution":{"iopub.execute_input":"2022-08-15T14:12:40.18029Z","iopub.status.busy":"2022-08-15T14:12:40.179358Z","iopub.status.idle":"2022-08-15T14:12:40.19118Z","shell.execute_reply":"2022-08-15T14:12:40.190217Z"},"lines_to_next_cell":0,"papermill":{"duration":0.01876,"end_time":"2022-08-15T14:12:40.193087","exception":false,"start_time":"2022-08-15T14:12:40.174327","status":"completed"},"tags":[]},"outputs":[{"data":{"application/javascript":["parent.postMessage({\"jupyterEvent\": \"custom.exercise_interaction\", \"data\": {\"outcomeType\": 1, \"valueTowardsCompletion\": 0.3333333333333333, \"interactionType\": 1, \"questionType\": 2, \"questionId\": \"2_Q2\", \"learnToolsVersion\": \"0.3.4\", \"failureMessage\": \"\", \"exceptionClass\": \"\", \"trace\": \"\"}}, \"*\")"],"text/plain":["<IPython.core.display.Javascript object>"]},"metadata":{},"output_type":"display_data"},{"data":{"text/markdown":["<span style=\"color:#33cc33\">Correct</span>"],"text/plain":["Correct"]},"metadata":{},"output_type":"display_data"}],"source":["model.compile(\n","    optimizer=tf.keras.optimizers.Adam(epsilon=0.01),\n","    loss='binary_crossentropy',\n","    metrics=['binary_accuracy'],\n",")\n","q_2.assert_check_passed()"]},{"cell_type":"code","execution_count":6,"id":"1bbe751e","metadata":{"execution":{"iopub.execute_input":"2022-08-15T14:12:40.203253Z","iopub.status.busy":"2022-08-15T14:12:40.202997Z","iopub.status.idle":"2022-08-15T14:12:40.206616Z","shell.execute_reply":"2022-08-15T14:12:40.205689Z"},"lines_to_next_cell":0,"papermill":{"duration":0.011005,"end_time":"2022-08-15T14:12:40.208529","exception":false,"start_time":"2022-08-15T14:12:40.197524","status":"completed"},"tags":[]},"outputs":[],"source":["# Lines below will give you a hint or solution code\n","#q_2.hint()\n","#q_2.solution()"]},{"cell_type":"markdown","id":"d8afde4e","metadata":{"papermill":{"duration":0.004253,"end_time":"2022-08-15T14:12:40.217367","exception":false,"start_time":"2022-08-15T14:12:40.213114","status":"completed"},"tags":[]},"source":["Finally, let's test the performance of this new model. First run this cell to fit the model to the training set."]},{"cell_type":"code","execution_count":7,"id":"4780b8d1","metadata":{"execution":{"iopub.execute_input":"2022-08-15T14:12:40.227842Z","iopub.status.busy":"2022-08-15T14:12:40.226996Z","iopub.status.idle":"2022-08-15T14:16:52.746223Z","shell.execute_reply":"2022-08-15T14:16:52.745255Z"},"papermill":{"duration":252.526616,"end_time":"2022-08-15T14:16:52.748349","exception":false,"start_time":"2022-08-15T14:12:40.221733","status":"completed"},"tags":[]},"outputs":[{"name":"stderr","output_type":"stream","text":["2022-08-15 14:12:40.266701: I tensorflow/compiler/mlir/mlir_graph_optimization_pass.cc:185] None of the MLIR Optimization Passes are enabled (registered 2)\n"]},{"name":"stdout","output_type":"stream","text":["Epoch 1/50\n"]},{"name":"stderr","output_type":"stream","text":["2022-08-15 14:12:43.133771: I tensorflow/stream_executor/cuda/cuda_dnn.cc:369] Loaded cuDNN version 8005\n"]},{"name":"stdout","output_type":"stream","text":["80/80 [==============================] - 33s 302ms/step - loss: 0.6816 - binary_accuracy: 0.5744 - val_loss: 0.6691 - val_binary_accuracy: 0.5785\n","Epoch 2/50\n","80/80 [==============================] - 3s 43ms/step - loss: 0.6667 - binary_accuracy: 0.5787 - val_loss: 0.6610 - val_binary_accuracy: 0.5785\n","Epoch 3/50\n","80/80 [==============================] - 3s 42ms/step - loss: 0.6621 - binary_accuracy: 0.5787 - val_loss: 0.6507 - val_binary_accuracy: 0.5785\n","Epoch 4/50\n","80/80 [==============================] - 3s 43ms/step - loss: 0.6534 - binary_accuracy: 0.5787 - val_loss: 0.6450 - val_binary_accuracy: 0.5785\n","Epoch 5/50\n","80/80 [==============================] - 3s 42ms/step - loss: 0.6493 - binary_accuracy: 0.6002 - val_loss: 0.6381 - val_binary_accuracy: 0.5991\n","Epoch 6/50\n","80/80 [==============================] - 3s 42ms/step - loss: 0.6390 - binary_accuracy: 0.6316 - val_loss: 0.6346 - val_binary_accuracy: 0.5955\n","Epoch 7/50\n","80/80 [==============================] - 3s 42ms/step - loss: 0.6371 - binary_accuracy: 0.6344 - val_loss: 0.6367 - val_binary_accuracy: 0.6023\n","Epoch 8/50\n","80/80 [==============================] - 3s 42ms/step - loss: 0.6276 - binary_accuracy: 0.6547 - val_loss: 0.6194 - val_binary_accuracy: 0.6452\n","Epoch 9/50\n","80/80 [==============================] - 3s 44ms/step - loss: 0.6236 - binary_accuracy: 0.6598 - val_loss: 0.6175 - val_binary_accuracy: 0.6440\n","Epoch 10/50\n","80/80 [==============================] - 3s 42ms/step - loss: 0.6169 - binary_accuracy: 0.6691 - val_loss: 0.6091 - val_binary_accuracy: 0.6638\n","Epoch 11/50\n","80/80 [==============================] - 3s 42ms/step - loss: 0.6064 - binary_accuracy: 0.6826 - val_loss: 0.6039 - val_binary_accuracy: 0.6710\n","Epoch 12/50\n","80/80 [==============================] - 3s 42ms/step - loss: 0.6024 - binary_accuracy: 0.6873 - val_loss: 0.6019 - val_binary_accuracy: 0.6698\n","Epoch 13/50\n","80/80 [==============================] - 3s 42ms/step - loss: 0.5899 - binary_accuracy: 0.6959 - val_loss: 0.5929 - val_binary_accuracy: 0.6840\n","Epoch 14/50\n","80/80 [==============================] - 3s 43ms/step - loss: 0.5828 - binary_accuracy: 0.7059 - val_loss: 0.5990 - val_binary_accuracy: 0.6789\n","Epoch 15/50\n","80/80 [==============================] - 3s 42ms/step - loss: 0.5687 - binary_accuracy: 0.7286 - val_loss: 0.5747 - val_binary_accuracy: 0.7113\n","Epoch 16/50\n","80/80 [==============================] - 3s 42ms/step - loss: 0.5503 - binary_accuracy: 0.7370 - val_loss: 0.5535 - val_binary_accuracy: 0.7161\n","Epoch 17/50\n","80/80 [==============================] - 3s 42ms/step - loss: 0.5263 - binary_accuracy: 0.7452 - val_loss: 0.5308 - val_binary_accuracy: 0.7434\n","Epoch 18/50\n","80/80 [==============================] - 3s 42ms/step - loss: 0.4979 - binary_accuracy: 0.7608 - val_loss: 0.5163 - val_binary_accuracy: 0.7408\n","Epoch 19/50\n","80/80 [==============================] - 3s 42ms/step - loss: 0.4739 - binary_accuracy: 0.7813 - val_loss: 0.5060 - val_binary_accuracy: 0.7531\n","Epoch 20/50\n","80/80 [==============================] - 3s 42ms/step - loss: 0.4418 - binary_accuracy: 0.8036 - val_loss: 0.4887 - val_binary_accuracy: 0.7753\n","Epoch 21/50\n","80/80 [==============================] - 3s 42ms/step - loss: 0.4199 - binary_accuracy: 0.8163 - val_loss: 0.5053 - val_binary_accuracy: 0.7410\n","Epoch 22/50\n","80/80 [==============================] - 3s 42ms/step - loss: 0.3867 - binary_accuracy: 0.8319 - val_loss: 0.4326 - val_binary_accuracy: 0.8052\n","Epoch 23/50\n","80/80 [==============================] - 3s 42ms/step - loss: 0.3524 - binary_accuracy: 0.8507 - val_loss: 0.4820 - val_binary_accuracy: 0.7612\n","Epoch 24/50\n","80/80 [==============================] - 3s 42ms/step - loss: 0.3386 - binary_accuracy: 0.8554 - val_loss: 0.4324 - val_binary_accuracy: 0.7957\n","Epoch 25/50\n","80/80 [==============================] - 3s 42ms/step - loss: 0.3089 - binary_accuracy: 0.8697 - val_loss: 0.4353 - val_binary_accuracy: 0.7886\n","Epoch 26/50\n","80/80 [==============================] - 3s 42ms/step - loss: 0.2793 - binary_accuracy: 0.8865 - val_loss: 0.4330 - val_binary_accuracy: 0.8101\n","Epoch 27/50\n","80/80 [==============================] - 3s 43ms/step - loss: 0.2638 - binary_accuracy: 0.8988 - val_loss: 0.4213 - val_binary_accuracy: 0.8236\n","Epoch 28/50\n","80/80 [==============================] - 3s 42ms/step - loss: 0.2580 - binary_accuracy: 0.8943 - val_loss: 0.4508 - val_binary_accuracy: 0.8240\n","Epoch 29/50\n","80/80 [==============================] - 3s 43ms/step - loss: 0.2387 - binary_accuracy: 0.9021 - val_loss: 0.4321 - val_binary_accuracy: 0.8157\n","Epoch 30/50\n","80/80 [==============================] - 3s 42ms/step - loss: 0.2155 - binary_accuracy: 0.9162 - val_loss: 0.4086 - val_binary_accuracy: 0.8379\n","Epoch 31/50\n","80/80 [==============================] - 3s 43ms/step - loss: 0.2069 - binary_accuracy: 0.9214 - val_loss: 0.4341 - val_binary_accuracy: 0.8266\n","Epoch 32/50\n","80/80 [==============================] - 3s 43ms/step - loss: 0.2185 - binary_accuracy: 0.9111 - val_loss: 0.4584 - val_binary_accuracy: 0.8293\n","Epoch 33/50\n","80/80 [==============================] - 3s 42ms/step - loss: 0.2165 - binary_accuracy: 0.9107 - val_loss: 0.6101 - val_binary_accuracy: 0.8048\n","Epoch 34/50\n","80/80 [==============================] - 3s 42ms/step - loss: 0.1811 - binary_accuracy: 0.9306 - val_loss: 0.6718 - val_binary_accuracy: 0.8072\n","Epoch 35/50\n","80/80 [==============================] - 3s 42ms/step - loss: 0.1589 - binary_accuracy: 0.9425 - val_loss: 0.6969 - val_binary_accuracy: 0.8054\n","Epoch 36/50\n","80/80 [==============================] - 3s 42ms/step - loss: 0.1445 - binary_accuracy: 0.9492 - val_loss: 0.6911 - val_binary_accuracy: 0.8086\n","Epoch 37/50\n","80/80 [==============================] - 3s 43ms/step - loss: 0.1331 - binary_accuracy: 0.9556 - val_loss: 0.7400 - val_binary_accuracy: 0.8050\n","Epoch 38/50\n","80/80 [==============================] - 3s 42ms/step - loss: 0.1316 - binary_accuracy: 0.9545 - val_loss: 0.5597 - val_binary_accuracy: 0.8373\n","Epoch 39/50\n","80/80 [==============================] - 3s 43ms/step - loss: 0.1397 - binary_accuracy: 0.9519 - val_loss: 0.5774 - val_binary_accuracy: 0.8420\n","Epoch 40/50\n","80/80 [==============================] - 4s 44ms/step - loss: 0.1258 - binary_accuracy: 0.9566 - val_loss: 0.5190 - val_binary_accuracy: 0.8390\n","Epoch 41/50\n","80/80 [==============================] - 3s 42ms/step - loss: 0.1123 - binary_accuracy: 0.9635 - val_loss: 0.5814 - val_binary_accuracy: 0.8442\n","Epoch 42/50\n","80/80 [==============================] - 3s 43ms/step - loss: 0.0997 - binary_accuracy: 0.9707 - val_loss: 0.5954 - val_binary_accuracy: 0.8357\n","Epoch 43/50\n","80/80 [==============================] - 3s 42ms/step - loss: 0.0983 - binary_accuracy: 0.9683 - val_loss: 0.5909 - val_binary_accuracy: 0.8299\n","Epoch 44/50\n","80/80 [==============================] - 3s 42ms/step - loss: 0.0973 - binary_accuracy: 0.9703 - val_loss: 0.5616 - val_binary_accuracy: 0.8149\n","Epoch 45/50\n","80/80 [==============================] - 3s 43ms/step - loss: 0.0878 - binary_accuracy: 0.9724 - val_loss: 0.6567 - val_binary_accuracy: 0.8070\n","Epoch 46/50\n","80/80 [==============================] - 3s 42ms/step - loss: 0.0812 - binary_accuracy: 0.9775 - val_loss: 0.5631 - val_binary_accuracy: 0.8388\n","Epoch 47/50\n","80/80 [==============================] - 3s 43ms/step - loss: 0.0699 - binary_accuracy: 0.9803 - val_loss: 0.6587 - val_binary_accuracy: 0.8505\n","Epoch 48/50\n","80/80 [==============================] - 3s 44ms/step - loss: 0.0716 - binary_accuracy: 0.9808 - val_loss: 0.6274 - val_binary_accuracy: 0.8462\n","Epoch 49/50\n","80/80 [==============================] - 3s 42ms/step - loss: 0.0693 - binary_accuracy: 0.9818 - val_loss: 0.6884 - val_binary_accuracy: 0.8462\n","Epoch 50/50\n","80/80 [==============================] - 3s 43ms/step - loss: 0.0622 - binary_accuracy: 0.9838 - val_loss: 0.6890 - val_binary_accuracy: 0.8501\n"]}],"source":["history = model.fit(\n","    ds_train,\n","    validation_data=ds_valid,\n","    epochs=50,\n",")"]},{"cell_type":"markdown","id":"e8103373","metadata":{"papermill":{"duration":0.108994,"end_time":"2022-08-15T14:16:52.967214","exception":false,"start_time":"2022-08-15T14:16:52.85822","status":"completed"},"tags":[]},"source":["And now run the cell below to plot the loss and metric curves for this training run."]},{"cell_type":"code","execution_count":8,"id":"377d534e","metadata":{"execution":{"iopub.execute_input":"2022-08-15T14:16:53.18564Z","iopub.status.busy":"2022-08-15T14:16:53.185288Z","iopub.status.idle":"2022-08-15T14:16:53.714527Z","shell.execute_reply":"2022-08-15T14:16:53.7136Z"},"papermill":{"duration":0.640869,"end_time":"2022-08-15T14:16:53.71662","exception":false,"start_time":"2022-08-15T14:16:53.075751","status":"completed"},"tags":[]},"outputs":[{"data":{"image/png":"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\n","text/plain":["<Figure size 432x288 with 1 Axes>"]},"metadata":{"needs_background":"light"},"output_type":"display_data"},{"data":{"image/png":"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\n","text/plain":["<Figure size 432x288 with 1 Axes>"]},"metadata":{"needs_background":"light"},"output_type":"display_data"}],"source":["import pandas as pd\n","history_frame = pd.DataFrame(history.history)\n","history_frame.loc[:, ['loss', 'val_loss']].plot()\n","history_frame.loc[:, ['binary_accuracy', 'val_binary_accuracy']].plot();"]},{"cell_type":"markdown","id":"406b15fd","metadata":{"papermill":{"duration":0.109843,"end_time":"2022-08-15T14:16:53.992116","exception":false,"start_time":"2022-08-15T14:16:53.882273","status":"completed"},"tags":[]},"source":["# 3) Train the Model #\n","\n","How would you interpret these training curves? Did this model improve upon the model from the tutorial?"]},{"cell_type":"code","execution_count":9,"id":"1fa677f6","metadata":{"execution":{"iopub.execute_input":"2022-08-15T14:16:54.212846Z","iopub.status.busy":"2022-08-15T14:16:54.212466Z","iopub.status.idle":"2022-08-15T14:16:54.221223Z","shell.execute_reply":"2022-08-15T14:16:54.220104Z"},"lines_to_next_cell":0,"papermill":{"duration":0.124386,"end_time":"2022-08-15T14:16:54.225116","exception":false,"start_time":"2022-08-15T14:16:54.10073","status":"completed"},"tags":[]},"outputs":[{"data":{"application/javascript":["parent.postMessage({\"jupyterEvent\": \"custom.exercise_interaction\", \"data\": {\"outcomeType\": 1, \"valueTowardsCompletion\": 0.3333333333333333, \"interactionType\": 1, \"questionType\": 4, \"questionId\": \"3_Q3\", \"learnToolsVersion\": \"0.3.4\", \"failureMessage\": \"\", \"exceptionClass\": \"\", \"trace\": \"\"}}, \"*\")"],"text/plain":["<IPython.core.display.Javascript object>"]},"metadata":{},"output_type":"display_data"},{"data":{"text/markdown":["<span style=\"color:#33cc33\">Correct:</span> \n","\n","\n","The learning curves for the model from the tutorial diverged fairly rapidly. This would indicate that it was prone to overfitting and in need of some regularization. The additional layer in our new model would make it even more prone to overfitting. However, adding some regularization with the `Dropout` layer helped prevent this. These changes improved the validation accuracy of the model by several points.\n"],"text/plain":["Correct: \n","\n","\n","The learning curves for the model from the tutorial diverged fairly rapidly. This would indicate that it was prone to overfitting and in need of some regularization. The additional layer in our new model would make it even more prone to overfitting. However, adding some regularization with the `Dropout` layer helped prevent this. These changes improved the validation accuracy of the model by several points."]},"metadata":{},"output_type":"display_data"}],"source":["# View the solution (Run this code cell to receive credit!)\n","q_3.check()"]},{"cell_type":"markdown","id":"be97a82b","metadata":{"papermill":{"duration":0.109886,"end_time":"2022-08-15T14:16:54.442468","exception":false,"start_time":"2022-08-15T14:16:54.332582","status":"completed"},"tags":[]},"source":["# Conclusion #\n","\n","These exercises showed you how to design a custom convolutional network to solve a specific classification problem. Though most models these days will be built on top of a pretrained base, it certain circumstances a smaller custom convnet might still be preferable -- such as with a smaller or unusual dataset or when computing resources are very limited. As you saw here, for certain problems they can perform just as well as a pretrained model.\n","\n","# Keep Going #\n","\n","Continue on to [**Lesson 6**](https://www.kaggle.com/ryanholbrook/data-augmentation), where you'll learn a widely-used technique that can give a boost to your training data: **data augmentation**."]},{"cell_type":"markdown","id":"2f6a00d9","metadata":{"papermill":{"duration":0.110722,"end_time":"2022-08-15T14:16:54.662809","exception":false,"start_time":"2022-08-15T14:16:54.552087","status":"completed"},"tags":[]},"source":["---\n","\n","\n","\n","\n","*Have questions or comments? Visit the [course discussion forum](https://www.kaggle.com/learn/computer-vision/discussion) to chat with other learners.*"]}],"metadata":{"kernelspec":{"display_name":"Python 3","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":"3.7.12"},"papermill":{"default_parameters":{},"duration":280.542886,"end_time":"2022-08-15T14:16:58.135871","environment_variables":{},"exception":null,"input_path":"__notebook__.ipynb","output_path":"__notebook__.ipynb","parameters":{},"start_time":"2022-08-15T14:12:17.592985","version":"2.3.4"}},"nbformat":4,"nbformat_minor":5}