v2_IAPPT.ipynb

{"cells":[{"cell_type":"markdown","metadata":{"id":"zFJvYZyHmZWN"},"source":["### IAPPT\n","Este archivo contiene la creacion y entrenamiento del modelo que usamos para determinar si la mano esta en posicion de piedra, papel o tijera.\n","Antes de nada ejecutaremos los imports necesarios."]},{"cell_type":"code","execution_count":118,"metadata":{"executionInfo":{"elapsed":256,"status":"ok","timestamp":1650536912091,"user":{"displayName":"Mario Ezquerro Ruiz","userId":"00902007422643275773"},"user_tz":-120},"id":"BV98OoRe06lB"},"outputs":[],"source":["import tensorflow as tf\n","import common.constants as const\n","import pandas as pd\n","\n","# Keras \n","from tensorflow.keras.utils import to_categorical\n","from keras.models import Sequential\n","from keras.layers import Dense\n"]},{"cell_type":"markdown","metadata":{"id":"GVRy1QM0mZWS"},"source":["### Datasets\n","Primero hemos de entender que datos estamos pasando a nuestro modelo exactamente.\n","Mediante mediapie obtenemos una serie de puntos llamados landmarks, cada uno con un numero representandolo. \n","\n","![alt text](https://google.github.io/mediapipe/images/mobile/hand_landmarks.png)\n","\n","Para obtener y guardar estos datos en el csv se encarga el ejecutable ```csv_generator.py```.\n","De todos estos landmarks solo usaremos estos: 0,6,8,10,12,14,16,18,20."]},{"cell_type":"code","execution_count":119,"metadata":{"executionInfo":{"elapsed":591,"status":"ok","timestamp":1650537676504,"user":{"displayName":"Mario Ezquerro Ruiz","userId":"00902007422643275773"},"user_tz":-120},"id":"KiYoFr_S0uhi"},"outputs":[{"data":{"text/html":["<div>\n","<style scoped>\n","    .dataframe tbody tr th:only-of-type {\n","        vertical-align: middle;\n","    }\n","\n","    .dataframe tbody tr th {\n","        vertical-align: top;\n","    }\n","\n","    .dataframe thead th {\n","        text-align: right;\n","    }\n","</style>\n","<table border=\"1\" class=\"dataframe\">\n","  <thead>\n","    <tr style=\"text-align: right;\">\n","      <th></th>\n","      <th>GESTO</th>\n","      <th>WRIST_Y</th>\n","      <th>WRIST_X</th>\n","      <th>INDEX_FINGER_PIP_Y</th>\n","      <th>INDEX_FINGER_PIP_X</th>\n","      <th>INDEX_FINGER_TIP_Y</th>\n","      <th>INDEX_FINGER_TIP_X</th>\n","      <th>MIDDLE_FINGER_PIP_Y</th>\n","      <th>MIDDLE_FINGER_PIP_X</th>\n","      <th>MIDDLE_FINGER_TIP_Y</th>\n","      <th>MIDDLE_FINGER_TIP_X</th>\n","      <th>RING_FINGER_PIP_Y</th>\n","      <th>RING_FINGER_PIP_X</th>\n","      <th>RING_FINGER_TIP_Y</th>\n","      <th>RING_FINGER_TIP_X</th>\n","      <th>PINKY_FINGER_PIP_Y</th>\n","      <th>PINKY_FINGER_PIP_X</th>\n","      <th>PINKY_FINGER_TIP_Y</th>\n","      <th>PINKY_FINGER_TIP_X</th>\n","    </tr>\n","  </thead>\n","  <tbody>\n","    <tr>\n","      <th>729</th>\n","      <td>1</td>\n","      <td>0.832705</td>\n","      <td>0.435459</td>\n","      <td>0.352794</td>\n","      <td>0.607703</td>\n","      <td>0.112611</td>\n","      <td>0.624792</td>\n","      <td>0.352675</td>\n","      <td>0.503303</td>\n","      <td>0.098365</td>\n","      <td>0.535860</td>\n","      <td>0.404979</td>\n","      <td>0.393161</td>\n","      <td>0.192138</td>\n","      <td>0.432784</td>\n","      <td>0.486190</td>\n","      <td>0.281573</td>\n","      <td>0.327257</td>\n","      <td>0.313231</td>\n","    </tr>\n","    <tr>\n","      <th>2029</th>\n","      <td>0</td>\n","      <td>0.634258</td>\n","      <td>0.475262</td>\n","      <td>0.349925</td>\n","      <td>0.604041</td>\n","      <td>0.291907</td>\n","      <td>0.613099</td>\n","      <td>0.329778</td>\n","      <td>0.541373</td>\n","      <td>0.295049</td>\n","      <td>0.555251</td>\n","      <td>0.338279</td>\n","      <td>0.470331</td>\n","      <td>0.298214</td>\n","      <td>0.485929</td>\n","      <td>0.375954</td>\n","      <td>0.397393</td>\n","      <td>0.341286</td>\n","      <td>0.414339</td>\n","    </tr>\n","    <tr>\n","      <th>1069</th>\n","      <td>2</td>\n","      <td>0.814044</td>\n","      <td>0.435166</td>\n","      <td>0.317014</td>\n","      <td>0.615717</td>\n","      <td>0.136898</td>\n","      <td>0.643111</td>\n","      <td>0.288134</td>\n","      <td>0.464042</td>\n","      <td>0.065704</td>\n","      <td>0.445972</td>\n","      <td>0.446139</td>\n","      <td>0.368890</td>\n","      <td>0.562781</td>\n","      <td>0.403117</td>\n","      <td>0.510620</td>\n","      <td>0.303959</td>\n","      <td>0.607387</td>\n","      <td>0.340711</td>\n","    </tr>\n","    <tr>\n","      <th>2268</th>\n","      <td>0</td>\n","      <td>0.718776</td>\n","      <td>0.429148</td>\n","      <td>0.336745</td>\n","      <td>0.557805</td>\n","      <td>0.334710</td>\n","      <td>0.539876</td>\n","      <td>0.320029</td>\n","      <td>0.466269</td>\n","      <td>0.337954</td>\n","      <td>0.475121</td>\n","      <td>0.340583</td>\n","      <td>0.381357</td>\n","      <td>0.360657</td>\n","      <td>0.410415</td>\n","      <td>0.401461</td>\n","      <td>0.304105</td>\n","      <td>0.406342</td>\n","      <td>0.342156</td>\n","    </tr>\n","    <tr>\n","      <th>202</th>\n","      <td>1</td>\n","      <td>0.771809</td>\n","      <td>0.474900</td>\n","      <td>0.367415</td>\n","      <td>0.745598</td>\n","      <td>0.169151</td>\n","      <td>0.852461</td>\n","      <td>0.304699</td>\n","      <td>0.585697</td>\n","      <td>0.079855</td>\n","      <td>0.641463</td>\n","      <td>0.331895</td>\n","      <td>0.456014</td>\n","      <td>0.106073</td>\n","      <td>0.474304</td>\n","      <td>0.370099</td>\n","      <td>0.376928</td>\n","      <td>0.205928</td>\n","      <td>0.369053</td>\n","    </tr>\n","  </tbody>\n","</table>\n","</div>"],"text/plain":["      GESTO   WRIST_Y   WRIST_X  INDEX_FINGER_PIP_Y  INDEX_FINGER_PIP_X  \\\n","729       1  0.832705  0.435459            0.352794            0.607703   \n","2029      0  0.634258  0.475262            0.349925            0.604041   \n","1069      2  0.814044  0.435166            0.317014            0.615717   \n","2268      0  0.718776  0.429148            0.336745            0.557805   \n","202       1  0.771809  0.474900            0.367415            0.745598   \n","\n","      INDEX_FINGER_TIP_Y  INDEX_FINGER_TIP_X  MIDDLE_FINGER_PIP_Y  \\\n","729             0.112611            0.624792             0.352675   \n","2029            0.291907            0.613099             0.329778   \n","1069            0.136898            0.643111             0.288134   \n","2268            0.334710            0.539876             0.320029   \n","202             0.169151            0.852461             0.304699   \n","\n","      MIDDLE_FINGER_PIP_X  MIDDLE_FINGER_TIP_Y  MIDDLE_FINGER_TIP_X  \\\n","729              0.503303             0.098365             0.535860   \n","2029             0.541373             0.295049             0.555251   \n","1069             0.464042             0.065704             0.445972   \n","2268             0.466269             0.337954             0.475121   \n","202              0.585697             0.079855             0.641463   \n","\n","      RING_FINGER_PIP_Y  RING_FINGER_PIP_X  RING_FINGER_TIP_Y  \\\n","729            0.404979           0.393161           0.192138   \n","2029           0.338279           0.470331           0.298214   \n","1069           0.446139           0.368890           0.562781   \n","2268           0.340583           0.381357           0.360657   \n","202            0.331895           0.456014           0.106073   \n","\n","      RING_FINGER_TIP_X  PINKY_FINGER_PIP_Y  PINKY_FINGER_PIP_X  \\\n","729            0.432784            0.486190            0.281573   \n","2029           0.485929            0.375954            0.397393   \n","1069           0.403117            0.510620            0.303959   \n","2268           0.410415            0.401461            0.304105   \n","202            0.474304            0.370099            0.376928   \n","\n","      PINKY_FINGER_TIP_Y  PINKY_FINGER_TIP_X  \n","729             0.327257            0.313231  \n","2029            0.341286            0.414339  \n","1069            0.607387            0.340711  \n","2268            0.406342            0.342156  \n","202             0.205928            0.369053  "]},"execution_count":119,"metadata":{},"output_type":"execute_result"}],"source":["# A almacenamos los datos de los ficheros\n","data_train = pd.read_csv(const.TRAINING_CSV)\n","data_eval = pd.read_csv(const.EVALUATION_CSV)\n","data_train.sample(5)"]},{"cell_type":"markdown","metadata":{"id":"T6youW8-qFyR"},"source":["Antes de usar el dataset eliminamos la columna que representa el resultado (Y) y la almacenamos en una variable.\n","\n"]},{"cell_type":"code","execution_count":120,"metadata":{"executionInfo":{"elapsed":256,"status":"ok","timestamp":1650537679598,"user":{"displayName":"Mario Ezquerro Ruiz","userId":"00902007422643275773"},"user_tz":-120},"id":"rlEKvIwx99bE"},"outputs":[],"source":["train_y = to_categorical(data_train.pop('GESTO'))\n","eval_y = to_categorical(data_eval.pop('GESTO'))"]},{"cell_type":"markdown","metadata":{},"source":["###  Creando el modelo\n","Usaremos un modelo secuencial con cuatro capas ```1 input``` (la cual se crea de manera implicita), ```2 ocultas``` y ```1 output``` con \n","Los parametros son estos:\n","\n","- ```20/10/3``` representa la cantidad de nodos de esta capa.\n","\n","- ```input_dim``` cantidad de parametros en la entrada.\n","\n","- ```activation``` una funcion de activacion opera los parametros del nodo y determina cual de los siguientes nodos se activa, este parametro especifica que tipo de funcion usamos.\n","\n","- ```optimizer``` los optimizadores son algoritmos usados para disminuir errores del modelo.\n","\n","- ```loss``` funcion utilizada para cuantificar la incertidumbre de la prediccion, cuanto mas cercano es a 0 mejor.\n","\n","- ```metrics``` es el objetivo del modelo (precision en este caso). "]},{"cell_type":"code","execution_count":121,"metadata":{"colab":{"base_uri":"https://localhost:8080/","height":552},"executionInfo":{"elapsed":260,"status":"error","timestamp":1650539868757,"user":{"displayName":"Mario Ezquerro Ruiz","userId":"00902007422643275773"},"user_tz":-120},"id":"SX_GWSBuOI-_","outputId":"5167ecfc-374f-4ff5-fd6f-2145c7f0957a"},"outputs":[{"name":"stdout","output_type":"stream","text":["Model: \"sequential\"\n","_________________________________________________________________\n"," Layer (type)                Output Shape              Param #   \n","=================================================================\n"," dense (Dense)               (None, 20)                380       \n","                                                                 \n"," dense_1 (Dense)             (None, 10)                210       \n","                                                                 \n"," dense_2 (Dense)             (None, 3)                 33        \n","                                                                 \n","=================================================================\n","Total params: 623\n","Trainable params: 623\n","Non-trainable params: 0\n","_________________________________________________________________\n"]}],"source":["tf.keras.backend.clear_session()\n","\n","model = Sequential()\n","model.add(Dense(20, activation='relu', input_dim=18))\n","model.add(Dense(10, activation='relu'))\n","model.add(Dense(3, activation='softmax'))\n","\n","# Compilar el modelo\n","model.compile(optimizer='Adam', \n","              loss='categorical_crossentropy', \n","              metrics=['accuracy'])\n","\n","# Mostrar estructura del modelo\n","model.summary()"]},{"cell_type":"markdown","metadata":{"id":"4j0Uf35VvcI7"},"source":["### Entrenar el modelo"]},{"cell_type":"markdown","metadata":{},"source":["Aqui es donde empieza a aprender y clasificar.\n","Para ello se llama con el metodo ```fit()``` con los siguientes parametros:\n","\n","- ```data_train``` es el dataset con los landmarks.\n","\n","- ```train_y``` contiene la columna con el dato de gesto para que el modelo compruebe si acierta o falla.\n","\n","- ```batch_size``` cantidad de filas de data_train que se procesan en cada paso.\n","\n","- ```epochs``` numero de iteraciones sobre el dataset completo."]},{"cell_type":"code","execution_count":122,"metadata":{"colab":{"base_uri":"https://localhost:8080/"},"executionInfo":{"elapsed":44085,"status":"ok","timestamp":1650539290192,"user":{"displayName":"Mario Ezquerro Ruiz","userId":"00902007422643275773"},"user_tz":-120},"id":"ndiyfeYcdwff","outputId":"81334d7a-d44e-48d1-eb48-8e9d2ece8ccf"},"outputs":[{"name":"stdout","output_type":"stream","text":["Epoch 1/10\n","72/72 [==============================] - 0s 693us/step - loss: 1.0992 - accuracy: 0.4148\n","Epoch 2/10\n","72/72 [==============================] - 0s 793us/step - loss: 1.0356 - accuracy: 0.6577\n","Epoch 3/10\n","72/72 [==============================] - 0s 732us/step - loss: 0.9432 - accuracy: 0.8779\n","Epoch 4/10\n","72/72 [==============================] - 0s 718us/step - loss: 0.7781 - accuracy: 0.9284\n","Epoch 5/10\n","72/72 [==============================] - 0s 662us/step - loss: 0.5658 - accuracy: 0.9539\n","Epoch 6/10\n","72/72 [==============================] - 0s 676us/step - loss: 0.3865 - accuracy: 0.9613\n","Epoch 7/10\n","72/72 [==============================] - 0s 676us/step - loss: 0.2630 - accuracy: 0.9736\n","Epoch 8/10\n","72/72 [==============================] - 0s 704us/step - loss: 0.1906 - accuracy: 0.9780\n","Epoch 9/10\n","72/72 [==============================] - 0s 662us/step - loss: 0.1452 - accuracy: 0.9837\n","Epoch 10/10\n","72/72 [==============================] - 0s 676us/step - loss: 0.1173 - accuracy: 0.9890\n"]}],"source":["history = model.fit(data_train, train_y, batch_size=32, epochs=10)"]},{"cell_type":"markdown","metadata":{},"source":["### Desempeño del modelo durante entrenamiento"]},{"cell_type":"code","execution_count":123,"metadata":{},"outputs":[{"data":{"text/plain":["<matplotlib.legend.Legend at 0x1e6f7114160>"]},"execution_count":123,"metadata":{},"output_type":"execute_result"},{"data":{"image/png":"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","text/plain":["<Figure size 720x432 with 1 Axes>"]},"metadata":{"needs_background":"light"},"output_type":"display_data"}],"source":["import numpy as np\n","import matplotlib.pyplot as plt\n","\n","plt.figure(figsize=(10, 6))\n","plt.title('Model performance')\n","plt.plot(np.arange(1, 11), history.history['loss'], label='Loss', lw=3)\n","plt.plot(np.arange(1, 11), history.history['accuracy'], label='Precision', lw=3)\n","plt.xlabel(\"Epoch\")\n","plt.legend()"]},{"cell_type":"markdown","metadata":{},"source":["### Resultados\n","Para visualizar la precision del modelo mostrandole nuevos datos, ejecutamos esta celda."]},{"cell_type":"code","execution_count":124,"metadata":{"colab":{"base_uri":"https://localhost:8080/"},"executionInfo":{"elapsed":882,"status":"ok","timestamp":1650539291069,"user":{"displayName":"Mario Ezquerro Ruiz","userId":"00902007422643275773"},"user_tz":-120},"id":"PQBxMfqN7jwa","outputId":"77e3de09-4647-4995-bc5a-716ec73dd610"},"outputs":[{"name":"stdout","output_type":"stream","text":["12/12 [==============================] - 0s 819us/step - loss: 0.1166 - accuracy: 0.9890\n","Precision del modelo: 0.9890\n"]}],"source":["scores = model.evaluate(data_eval, eval_y)\n","print(f'Precision del modelo: {scores[1]:.4f}')  "]},{"cell_type":"markdown","metadata":{},"source":["### Guardar el modelo\n","Si estamos contentos con el rendimiento lo podemos guardar de esta manera:"]},{"cell_type":"code","execution_count":84,"metadata":{},"outputs":[],"source":["model.save('model\\\\gesture_classifier.h5')"]}],"metadata":{"colab":{"collapsed_sections":[],"name":"IAPPT.ipynb","provenance":[]},"interpreter":{"hash":"50ade791ad1c8b9f359d547d69534c43a84b1fbf6337af47cd451964d6c5b72f"},"kernelspec":{"display_name":"Python 3.8.10 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":"3.8.10"}},"nbformat":4,"nbformat_minor":0}