Merge remote-tracking branch 'origin/master'

docs/html/modeling.html

@@ -151,6 +151,14 @@ <h1 id="cq">convolutional_query()</h1>
                         <td>augmentation=True</td>
                         <td>Magnifies dataset to add small variation to images before training in the model.</td>
                     </tr>
+                    <tr>
+                        <td>custom_arch=None</td>
+                        <td>File path to a custom Keras model architecture configuration in JSON format. </td>
+                    </tr>
+                    <tr>
+                        <td>pretrained=None</td>
+                        <td>Load a state-of-the-art model architecture through dictionary with keys 'arch' and 'weights'. Current values supported for 'arch' key are 'vggnet16', 'vggnet19', 'resnet50', 'resnet101', 'resnet152'. Possible values for 'weights' key include 'imagenet' and 'random'. </td>
+                    </tr>
                     <tr>
                         <td>epochs=10</td>
                         <td>Number of epochs for every model attempted.</td>

libra/queries.py

@@ -745,6 +745,7 @@ def convolutional_query(self,
                             read_mode=None,
                             verbose=0,
                             preprocess=True,
+                            data_path = None,
                             new_folders=True,
                             image_column=None,
                             test_size=0.2,

libra/query/feedforward_nn.py

@@ -494,7 +494,9 @@ def convolutional(instruction=None,
     :param many parameters: used to preprocess, tune, plot generation, and parameterizing the convolutional neural network trained.
     :return dictionary that holds all the information for the finished model.
     '''
-    # data_path = get_folder_dir()
+
+    #data_path = get_folder_dir()
+
     logger("Generating datasets for classes")
 
     if pretrained:
@@ -503,11 +505,11 @@ def convolutional(instruction=None,
         if not width:
             width = 224
         if height != 224 or width != 224:
-            raise ValueError("For pretrained models, height must be 224 and width must be 224.")
+            raise ValueError("For pretrained models, both 'height' and 'width' must be 224.")
 
     if preprocess:
         if custom_arch:
-            raise ValueError("If custom_arch is not None, preprocess must be set to false.")
+            raise ValueError("If 'custom_arch' is not None, 'preprocess' must be set to false.")
 
         read_mode_info = set_distinguisher(data_path, read_mode)
         read_mode = read_mode_info["read_mode"]
@@ -556,11 +558,15 @@ def convolutional(instruction=None,
     input_single = (processInfo["height"], processInfo["width"])
     num_classes = processInfo["num_categories"]
     loss_func = ""
+    output_layer_activation = ""
 
     if num_classes > 2:
         loss_func = "categorical_crossentropy"
+        output_layer_activation = "softmax"
     elif num_classes == 2:
+        num_classes = 1
         loss_func = "binary_crossentropy"
+        output_layer_activation = "sigmoid"
 
     logger("Creating convolutional neural netwwork dynamically")
 
@@ -588,7 +594,7 @@ def convolutional(instruction=None,
                 x = Dropout(0.5)(x)
                 x = Dense(4096)(x)
                 x = Dropout(0.5)(x)
-                pred = Dense(num_classes, activation='softmax')(x)
+                pred = Dense(num_classes, activation=output_layer_activation)(x)
                 model = Model(base_model.input, pred)
             elif arch_lower == "vggnet19":
                 base_model = VGG19(include_top=False, weights='imagenet', input_shape=input_shape)
@@ -597,36 +603,36 @@ def convolutional(instruction=None,
                 x = Dropout(0.5)(x)
                 x = Dense(4096)(x)
                 x = Dropout(0.5)(x)
-                pred = Dense(num_classes, activation='softmax')(x)
+                pred = Dense(num_classes, activation=output_layer_activation)(x)
                 model = Model(base_model.input, pred)
             elif arch_lower == "resnet50":
                 base_model = ResNet50(include_top=False, weights='imagenet', input_shape=input_shape)
                 x = Flatten()(base_model.output)
                 x = GlobalAveragePooling2D()(base_model.output)
                 x = Dropout(0.5)(x)
-                pred = Dense(num_classes, activation='softmax')(x)
+                pred = Dense(num_classes, activation=output_layer_activation)(x)
                 model = Model(base_model.input, pred)
             elif arch_lower == "resnet101":
                 base_model = ResNet101(include_top=False, weights='imagenet', input_shape=input_shape)
                 x = GlobalAveragePooling2D()(base_model.output)
                 x = Dropout(0.5)(x)
-                pred = Dense(num_classes, activation='softmax')(x)
+                pred = Dense(num_classes, activation=output_layer_activation)(x)
                 model = Model(base_model.input, pred)
             elif arch_lower == "resnet152":
                 base_model = ResNet152(include_top=False, weights='imagenet', input_shape=input_shape)
                 x = GlobalAveragePooling2D()(base_model.output)
                 x = Dropout(0.5)(x)
-                pred = Dense(num_classes, activation='softmax')(x)
+                pred = Dense(num_classes, activation=output_layer_activation)(x)
                 model = Model(base_model.input, pred)
             else:
                 raise ModuleNotFoundError("arch \'" + pretrained.get('arch') + "\' not supported.")
 
         else:
             # Randomly initialized weights
             if arch_lower == "vggnet16":
-                model = VGG16(include_top=True, weights=None, classes=num_classes)
+                model = VGG16(include_top=True, weights=None, classes=num_classes, classifier_activation = output_layer_activation)
             elif arch_lower == "vggnet19":
-                model = VGG19(include_top=True, weights=None, classes=num_classes)
+                model = VGG19(include_top=True, weights=None, classes=num_classes, classifier_activation = output_layer_activation)
             elif arch_lower == "resnet50":
                 model = ResNet50(include_top=True, weights=None, classes=num_classes)
             elif arch_lower == "resnet101":
@@ -679,6 +685,7 @@ def convolutional(instruction=None,
             activation="softmax"
         ))
 
+
     model.compile(
         optimizer="adam",
         loss=loss_func,
@@ -702,15 +709,15 @@ def convolutional(instruction=None,
     X_train = train_data.flow_from_directory(data_path + training_path,
                                              target_size=input_single,
                                              color_mode=color_mode,
-                                             batch_size=(32 if processInfo["train_size"] >= 32 else 1),
+                                             batch_size=(16 if processInfo["train_size"] >= 16 else 1),
                                              class_mode=loss_func[:loss_func.find("_")])
     X_test = test_data.flow_from_directory(data_path + testing_path,
                                            target_size=input_single,
                                            color_mode=color_mode,
-                                           batch_size=(32 if processInfo["test_size"] >= 32 else 1),
+                                           batch_size=(16 if processInfo["test_size"] >= 16 else 1),
                                            class_mode=loss_func[:loss_func.find("_")])
 
-    if epochs < 0:
+    if epochs <= 0:
         raise BaseException("Number of epochs has to be greater than 0.")
     logger('Training image model')
     history = model.fit_generator(

tests/tests.py

@@ -22,7 +22,6 @@ def compare(a, b):
 
 class TestQueries(unittest.TestCase):
     newClient = client('tools/data/structured_data/housing.csv')
-
     """
     TEST QUERIES
     
@@ -53,6 +52,35 @@ def test_nn_query(self):
         # see if properly chooses classification with a categorical target column
         self.newClient.neural_network_query('predict ocean proximity', epochs=3)
         self.assertTrue('classification_ANN' in self.newClient.models)
+    '''
+    @ordered
+    def test_convolutional_query(self):
+        client_image = client("tools/data/image_data/character_dataset_mini")
+        client_image.convolutional_query("predict character", epochs=2)
+        self.assertTrue('convolutional_NN' in client_image.models)
+    '''
+
+    @ordered
+    def test_convolutional_query_customarch(self):
+        data_path = "tools/data/image_data/character_dataset_mini_preprocessed"
+        client_image_customarch = client(data_path)
+        custom_arch_path = "tools/data/custom_model_config/custom_CNN.json"
+
+        client_image_customarch.convolutional_query("predict character", data_path = data_path, custom_arch=custom_arch_path, preprocess=False, epochs=2)
+        self.assertTrue('convolutional_NN' in client_image_customarch.models)
+
+
+    @ordered
+    def test_convolutional_query_pretrained(self):
+        client_image = client("tools/data/image_data/character_dataset_mini")
+        client_image.convolutional_query(
+            "predict character",
+            pretrained={
+                'arch': 'vggnet19',
+                'weights': 'imagenet'
+                },
+            epochs=2)
+        self.assertTrue('convolutional_NN' in client_image.models)
 
     # Tests whether decision_tree_query works without errors, and creates a key in models dictionary
     @ordered
@@ -117,6 +145,8 @@ def test_content_recommender(self):
         x.content_recommender_query()
         assert ('recommendations' in x.recommend('Coco'))
 
+
+
     """
     TEST ANALYZE() FUNCTION
     
@@ -173,6 +203,5 @@ def test_invalid_model(self):
         with self.assertRaises(NameError):
             self.newClient.analyze(model='I dont exist')
 
-
 if __name__ == '__main__':
     unittest.main()