first commit

house_prices_estimation_example_data.py

 # -*- coding: utf-8 -*-
-
+from tensorflow_core.python.keras.engine.input_layer import Input
 from tensorflow_core.python.keras.layers.core import Dense, Dropout
-from tensorflow_core.python.keras.models import Sequential
+from tensorflow_core.python.keras.models import Sequential, Model
 from tensorflow_core.python.keras.optimizer_v2.adam import Adam
 
 from datasets import load_house_dataset_data
@@ -10,13 +10,21 @@ import matplotlib.pyplot as plt
 
 __author__ = 106360
 
-def generate_simple_regression_model(input_shape):
-    model = Sequential()
-    model.add(Dense(16, input_dim=input_shape, activation="relu"))
-    # model.add(Dropout(0.25))
-    model.add(Dense(6, activation="relu"))
-    # model.add(Dropout(0.25))
-    model.add(Dense(1, activation="sigmoid"))
+def generate_simple_regression_model(input_shape,weights='',remove_head=False):
+    # define the model input
+    inputs = Input(shape=(input_shape))
+    # loop over the number of filters
+
+    x = Dense(16, input_dim=input_shape, activation="relu", name='layer1')(inputs)
+    x = Dense(6, activation="relu",name='layer2')(x)
+
+    if not remove_head:
+        x= Dense(1, activation="sigmoid",name='layer3')(x)
+
+    model = Model(inputs,x)
+    if weights!='':
+        model.load_weights(weights,by_name=True)
+
 
     return model
 

house_prices_estimation_example_image.py

@@ -11,13 +11,13 @@ from tensorflow_core.python.keras.models import Sequential, Model, load_model
 from tensorflow_core.python.keras.optimizer_v2.adam import Adam
 import tensorflow as tf
 
-from train_and_evaluation import evaluate_regression_model
+from train_and_evaluation import evaluate_regression_model, train_model
 from datasets import load_house_dataset_data, DatasetType
 import matplotlib.pyplot as plt
 
 __author__ = 106360
 
-def generate_simple_cnn_regression_model(input_shape,n_blocks=2,weights='',is_regression=True,num_classes=1,freeze=False):
+def generate_simple_cnn_regression_model(input_shape,n_blocks=2,weights='',is_regression=True,num_classes=1,freeze=False,remove_head=False):
     # define the model input
     inputs = Input(shape=(input_shape,input_shape,3))
     # loop over the number of filters
@@ -29,12 +29,14 @@ def generate_simple_cnn_regression_model(input_shape,n_blocks=2,weights='',is_re
         x = MaxPooling2D()(x)
 
     x= Flatten()(x)
-
+    x = Dense(6, activation="relu", name='pre_last_dense_reg')(x)
+    if not remove_head:
         if is_regression:
             y = Dense(num_classes, activation="sigmoid",name='last_dense_reg')(x)
         else:
             y = Dense(num_classes, activation="softmax", name='last_dense_clf')(x)
-
+    else:
+        y=x
     model = Model(inputs,y)
 
     if weights!='':
@@ -115,7 +117,7 @@ def train_cifar100(num_classes=100,batch_size=32):
 
 
 if __name__ == "__main__":
-    pre_train_with_cifar100 = False
+    pre_train_with_cifar100 = True
     weights = ''
     if pre_train_with_cifar100:
         file_weight_cifar100 = 'pretrained_cifar100.h5'

house_prices_estimation_example_image_and_data.py

 # -*- coding: utf-8 -*-
+from tensorflow_core.python.keras.layers.merge import Concatenate
+
+
+from house_prices_estimation_example_data import generate_simple_regression_model
+from house_prices_estimation_example_image import generate_simple_cnn_regression_model
+
+__author__ = 106360
 import os
 
 from tensorflow_core.python.keras.datasets import cifar10
@@ -11,80 +18,43 @@ from tensorflow_core.python.keras.models import Sequential, Model, load_model
 from tensorflow_core.python.keras.optimizer_v2.adam import Adam
 import tensorflow as tf
 
-from utils import plot_regression, load_house_dataset_data, DatasetType
+from datasets import load_house_dataset_data, DatasetType
 import matplotlib.pyplot as plt
 
-__author__ = 106360
-
-
-def train_model(trainX, trainY, testX, testY, model, epochs=200, batch_size=16,show_plot=True):
-    history = model.fit(trainX, trainY, validation_data=(testX, testY),
-                        epochs=epochs, batch_size=batch_size)
-
-    if show_plot:
-        acc = history.history['mean_absolute_percentage_error']
-        val_acc = history.history['val_mean_absolute_percentage_error']
-        loss = history.history['loss']
-        val_loss = history.history['val_loss']
-
-        epochs_range = range(epochs)
+from train_and_evaluation import train_model, evaluate_regression_model
 
-        plt.figure(figsize=(8, 8))
-        plt.subplot(1, 2, 1)
-        plt.plot(epochs_range, acc, label='Training mean_absolute_percentage_error')
-        plt.plot(epochs_range, val_acc, label='Validation mean_absolute_percentage_error')
-        plt.legend(loc='upper right')
-        plt.title('Training and Validation \% error')
 
-        plt.subplot(1, 2, 2)
-        plt.plot(epochs_range, loss, label='Training Loss')
-        plt.plot(epochs_range, val_loss, label='Validation Loss')
-        plt.legend(loc='upper right')
-        plt.title('Training and Validation Loss')
-        plt.show()
 
-    return model
 
 
+if __name__ == "__main__":
+    (trainX_data,trainX_img, trainY, testX_data,testX_img,testY), normalizer = load_house_dataset_data(test_size=0.2,random_state=666,type=DatasetType.Both)
 
 
-if __name__ == "__main__":
-    pre_train_with_cifar100 = False
-    weights = ''
-    if pre_train_with_cifar100:
-        file_weight_cifar100 = 'pretrained_cifar100.h5'
+    trainX = [trainX_data,trainX_img['frontal_img']]
+    testX = [testX_data,testX_img['frontal_img']]
+    input_shape_data = trainX[0].shape[1]
+    input_shape_img = trainX[1].shape[1]
 
-        try:
-            model = load_model(file_weight_cifar100)
-        except:
-            model = train_cifar100()
-            model.save(file_weight_cifar100)
-        weights = file_weight_cifar100
 
 
-    (trainX_data,trainX_img, trainY, testX_data,testX_img,testY), normalizer = load_house_dataset_data(test_size=0.2,random_state=666,type=DatasetType.Both)
+    model_data = generate_simple_regression_model(input_shape_data, weights='regression_model_data.h5',remove_head=True)
+    model_img = generate_simple_cnn_regression_model(input_shape_img, weights='regression_model_image_pretrained.h5',remove_head=False)
+    input_data = [Input(input_shape_data),Input((input_shape_img,input_shape_img,3))]
 
+    # y_data = model_data.layers[-2]#(input_data[0])
+    # y_img = model_img.layers[-2]#(input_data[1])
+    y_data = model_data(input_data[0])
+    y_img = model_img(input_data[1])
 
-    trainX = trainX_img['frontal_img']
-    testX = testX_img['frontal_img']
-    input_shape = trainX.shape[1]
+    y = Concatenate()([y_data, y_img])
+    y = Dense(1,activation='sigmoid')(y)
 
-    if pre_train_with_cifar100:
-        file_weight_finetune = 'regression_model_image_finetune.h5'
-        model = generate_simple_cnn_regression_model(input_shape,weights=weights,freeze=True)
-        opt = Adam(lr=1e-3, decay=1e-3 / 200)
-        model.compile(loss='mean_squared_error',metrics=['mean_absolute_percentage_error','mean_absolute_error','mean_squared_error'], optimizer=opt)
-        model.summary()
-        model = train_model(trainX, trainY, testX, testY,model,show_plot=True,epochs=500,batch_size=32)
-        evaluate_regression_model(model,testX,testY,normalizer,show_plot=True)
-        model.save(file_weight_finetune)
-        weights = file_weight_finetune
-        final_model_weight = 'regression_model_image_pretrained.h5'
-    else:
-        final_model_weight = 'regression_model_image_from_scratch.h5'
+    for layer in model_img.layers:
+        layer.trainable = False
+    model = Model(input_data,y)
 
 
-    model = generate_simple_cnn_regression_model(input_shape, weights=weights)
     opt = Adam(lr=1e-3, decay=1e-3 / 200)
     model.compile(loss='mean_squared_error',
                   metrics=['mean_absolute_percentage_error', 'mean_absolute_error', 'mean_squared_error'],
@@ -92,4 +62,4 @@ if __name__ == "__main__":
     model.summary()
     model = train_model(trainX, trainY, testX, testY, model, show_plot=True, epochs=500, batch_size=32)
     evaluate_regression_model(model, testX, testY, normalizer, show_plot=True)
-    model.save(final_model_weight)
+    model.save('regression_model_combined.h5')