Final Version MS2

.idea/other.xml

@@ -0,0 +1,6 @@
+<?xml version="1.0" encoding="UTF-8"?>
+<project version="4">
+  <component name="PySciProjectComponent">
+    <option name="PY_SCI_VIEW_SUGGESTED" value="true" />
+  </component>
+</project>

ConvolutionalNeuralNetwork.py

@@ -7,6 +7,8 @@ import seaborn as sns
 import pickle
 import random
 import os
+import numpy as np
+from keras.utils import np_utils
 
 with open(os.path.join("dataset", "train.p"), mode='rb') as training_data:
     train = pickle.load(training_data)
@@ -33,6 +35,10 @@ mask = np.isin(y_valid, range(20))
 X_valid_subset = X_valid_norm[mask]
 y_valid_subset = y_valid[mask]
 
+num_classes = 20  # Anzahl der Straßenschilder-Klassen
+y_train_subset = np_utils.to_categorical(y_train_subset, num_classes)
+y_valid_subset = np_utils.to_categorical(y_valid_subset, num_classes)
+
 
 
 from tensorflow.keras import datasets, layers, models
@@ -43,23 +49,24 @@ model = models.Sequential()
 model.add(layers.Conv2D( filters = 2 , kernel_size = ( 3 , 3 ), padding = "same" , activation = 'relu' , input_shape = ( 32 , 32 , 3)))
 
 # TODO: Add layers to the model:
-#model.add(layers.AveragePooling2D(pool_size=(2, 2),strides=(1, 1), padding='same'))
+model.add(layers.MaxPooling2D(pool_size=(2, 2)))  # Max-Pooling-Schicht
-#model.add(layers.MaxPool2D(pool_size=(2, 2), strides=None, padding='valid', data_format=None))
+model.add(layers.Conv2D(64, (3, 3), activation='relu', padding='same'))  # Zweite Faltungs- und Aktivierungsschicht
-#model.add(layers.Dropout(.2, input_shape=(2,)))
+model.add(layers.MaxPooling2D(pool_size=(2, 2)))  # Max-Pooling-Schicht
-model.add(layers.Flatten())
+model.add(layers.Flatten())  # Flachschicht
-#model.add(layers.Dense(43, activation='softmax'))
+model.add(layers.Dense(128, activation='relu'))  # Versteckte Schicht mit 128 Neuronen und ReLU-Aktivierung
+model.add(layers.Dense(20, activation='softmax'))  # Ausgabeschicht mit Anzahl der Klassen und Softmax-Aktivierung
 
 
 # Prints a summary of your network
 model.summary()
 
-model.compile(optimizer = 'Adam', loss = 'sparse_categorical_crossentropy', metrics = ['accuracy'])
+model.compile(optimizer = 'Adam', loss = 'categorical_crossentropy', metrics = ['accuracy'])
 
 # TODO: Choose the batch size and the epochs
 history = model.fit(x = X_train_subset,
                     y = y_train_subset,
                     batch_size = 32,
-                    epochs = 1000,
+                    epochs = 5,
                     verbose = 1,
                     validation_data = (X_valid_subset, y_valid_subset))
 

FullyConnectedNeuralNetwork.py

@@ -7,6 +7,8 @@ import seaborn as sns
 import pickle
 import random
 import os
+import numpy as np
+from keras.utils import np_utils
 
 with open(os.path.join("dataset", "train.p"), mode='rb') as training_data:
     train = pickle.load(training_data)
@@ -33,33 +35,31 @@ mask = np.isin(y_valid, range(20))
 X_valid_subset = X_valid_norm[mask]
 y_valid_subset = y_valid[mask]
 
+num_classes = 20  # Anzahl der Straßenschilder-Klassen
+y_train_subset = np_utils.to_categorical(y_train_subset, num_classes)
+y_valid_subset = np_utils.to_categorical(y_valid_subset, num_classes)
+
 
 
 from tensorflow.keras import datasets, layers, models
 model = models.Sequential()
 
-# Only in the first layer you have to select the input_shape of the data (image).
-# TODO: Replace the question marks:
-model.add(layers.Conv2D( filters = 2 , kernel_size = ( 3 , 3 ), padding = "same" , activation = 'relu' , input_shape = ( 32 , 32 , 3)))
-
 # TODO: Add layers to the model:
-model.add(layers.AveragePooling2D(pool_size=(2, 2),strides=(1, 1), padding='same'))
+model.add(layers.Flatten(input_shape=(32, 32, 3)))  # Eingabeschicht, Bildgröße: 32x32, 3 Kanäle (RGB)
-model.add(layers.MaxPool2D(pool_size=(2, 2), strides=None, padding='valid', data_format=None))
+model.add(layers.Dense(128, activation='relu'))  # Versteckte Schicht mit 128 Neuronen und ReLU-Aktivierung
-model.add(layers.Dropout(.2, input_shape=(2,)))
+model.add(layers.Dense(20, activation='softmax'))  # Ausgabeschicht mit Anzahl der Klassen und Softmax-Aktivierung
-model.add(layers.Flatten())
-model.add(layers.Dense(43, activation='softmax'))
 
 
 # Prints a summary of your network
 model.summary()
 
-model.compile(optimizer = 'Adam', loss = 'sparse_categorical_crossentropy', metrics = ['accuracy'])
+model.compile(optimizer = 'rmsprop', loss = 'categorical_crossentropy', metrics = ['accuracy'])
 
 # TODO: Choose the batch size and the epochs
 history = model.fit(x = X_train_subset,
                     y = y_train_subset,
                     batch_size = 32,
-                    epochs = 1500,
+                    epochs = 5,
                     verbose = 1,
                     validation_data = (X_valid_subset, y_valid_subset))
 

OwnModel.py

@@ -7,6 +7,8 @@ import seaborn as sns
 import pickle
 import random
 import os
+import numpy as np
+from keras.utils import np_utils
 
 with open(os.path.join("dataset", "train.p"), mode='rb') as training_data:
     train = pickle.load(training_data)
@@ -33,6 +35,10 @@ mask = np.isin(y_valid, range(20))
 X_valid_subset = X_valid_norm[mask]
 y_valid_subset = y_valid[mask]
 
+num_classes = 20  # Anzahl der Straßenschilder-Klassen
+y_train_subset = np_utils.to_categorical(y_train_subset, num_classes)
+y_valid_subset = np_utils.to_categorical(y_valid_subset, num_classes)
+
 
 
 from tensorflow.keras import datasets, layers, models
@@ -47,19 +53,19 @@ model.add(layers.AveragePooling2D(pool_size=(2, 2),strides=(1, 1), padding='same
 model.add(layers.MaxPool2D(pool_size=(2, 2), strides=None, padding='valid', data_format=None))
 model.add(layers.Dropout(.2, input_shape=(2,)))
 model.add(layers.Flatten())
-#model.add(layers.Dense(43, activation='softmax'))
+model.add(layers.Dense(20, activation='softmax'))
 
 
 # Prints a summary of your network
 model.summary()
 
-model.compile(optimizer = 'Adam', loss = 'sparse_categorical_crossentropy', metrics = ['accuracy'])
+model.compile(optimizer = 'Adam', loss = 'categorical_crossentropy', metrics = ['accuracy'])
 
 # TODO: Choose the batch size and the epochs
 history = model.fit(x = X_train_subset,
                     y = y_train_subset,
                     batch_size = 32,
-                    epochs = 1000,
+                    epochs = 500,
                     verbose = 1,
                     validation_data = (X_valid_subset, y_valid_subset))
 

TestMitTest.py

@@ -0,0 +1,53 @@
+#Import libaries and datasets
+import matplotlib.pyplot as plt
+import numpy as np
+import tensorflow as tf
+import pandas as pd
+import seaborn as sns
+import pickle
+import random
+import os
+import numpy as np
+from keras.utils import np_utils
+
+with open(os.path.join("test", "test.p"), mode='rb') as validation_data:
+    valid = pickle.load(validation_data)
+
+X_valid, y_valid = valid['features'], valid['labels']
+
+from sklearn.utils import shuffle
+X_valid, y_valid = shuffle(X_valid, y_valid)
+
+# Normalize image to [0, 1]
+X_valid_norm = X_valid / 255
+
+#Wähle Klassen 0-19
+mask = np.isin(y_valid, range(20))
+X_valid_subset = X_valid_norm[mask]
+y_valid_subset = y_valid[mask]
+
+num_classes = 20  # Anzahl der Straßenschilder-Klassen
+y_valid_subset = np_utils.to_categorical(y_valid_subset, num_classes)
+
+
+
+#ownModel
+#convolutionalNeuralNetwork
+#fullyConnectedNeuralNetwork
+model = tf.keras.models.load_model('saved_model/fullyConnectedNeuralNetwork.h5')
+
+
+
+score = model.evaluate(X_valid_subset, y_valid_subset)
+print('Test Accuracy: {}'.format(score[1]))
+
+from sklearn.metrics import confusion_matrix
+predicted_classes = np.argmax(model.predict(X_valid_subset), axis=-1)
+y_true = np.argmax(y_valid_subset, axis=-1)
+
+cm = confusion_matrix(y_true, predicted_classes)
+
+plt.figure(figsize = (25, 25))
+sns.heatmap(cm, annot = True)
+
+plt.show()

TestMitValid.py

@@ -7,21 +7,18 @@ import seaborn as sns
 import pickle
 import random
 import os
+import numpy as np
+from keras.utils import np_utils
 
-with open(os.path.join("dataset", "train.p"), mode='rb') as training_data:
-    train = pickle.load(training_data)
 with open(os.path.join("dataset", "valid.p"), mode='rb') as validation_data:
     valid = pickle.load(validation_data)
 
-X_train, y_train = train['features'], train['labels']
 X_valid, y_valid = valid['features'], valid['labels']
 
 from sklearn.utils import shuffle
-X_train, y_train = shuffle(X_train, y_train)
 X_valid, y_valid = shuffle(X_valid, y_valid)
 
 # Normalize image to [0, 1]
-X_train_norm = X_train / 255
 X_valid_norm = X_valid / 255
 
 #Wähle Klassen 0-19
@@ -29,6 +26,10 @@ mask = np.isin(y_valid, range(20))
 X_valid_subset = X_valid_norm[mask]
 y_valid_subset = y_valid[mask]
 
+num_classes = 20  # Anzahl der Straßenschilder-Klassen
+y_valid_subset = np_utils.to_categorical(y_valid_subset, num_classes)
+
+
 
 #ownModel
 #convolutionalNeuralNetwork
@@ -42,7 +43,7 @@ print('Test Accuracy: {}'.format(score[1]))
 
 from sklearn.metrics import confusion_matrix
 predicted_classes = np.argmax(model.predict(X_valid_subset), axis=-1)
-y_true = y_valid_subset
+y_true = np.argmax(y_valid_subset, axis=-1)
 
 cm = confusion_matrix(y_true, predicted_classes)
 

TestOnePictureTest.py

@@ -0,0 +1,54 @@
+#Import libaries and datasets
+import matplotlib.pyplot as plt
+import numpy as np
+import tensorflow as tf
+import pandas as pd
+import seaborn as sns
+import pickle
+import random
+import os
+import numpy as np
+from keras.utils import np_utils
+
+with open(os.path.join("test", "test.p"), mode='rb') as validation_data:
+    valid = pickle.load(validation_data)
+
+X_valid, y_valid = valid['features'], valid['labels']
+
+from sklearn.utils import shuffle
+X_valid, y_valid = shuffle(X_valid, y_valid)
+
+# Normalize image to [0, 1]
+X_valid_norm = X_valid / 255
+
+#TakeOnePicture (Klasse 0-19)
+filtered_indices = [i for i, label in enumerate(y_valid) if label >= 0 and label <= 19]
+indice = random.choice(filtered_indices)
+X_valid_subset = X_valid[indice][np.newaxis, ...]
+y_valid_subset = y_valid[indice][np.newaxis, ...]
+
+num_classes = 20  # Anzahl der Straßenschilder-Klassen
+y_valid_subset = np_utils.to_categorical(y_valid_subset, num_classes)
+
+
+
+#ownModel
+#convolutionalNeuralNetwork
+#fullyConnectedNeuralNetwork
+model = tf.keras.models.load_model('saved_model/fullyConnectedNeuralNetwork.h5')
+
+
+
+score = model.evaluate(X_valid_subset, y_valid_subset)
+print('Test Accuracy: {}'.format(score[1]))
+
+from sklearn.metrics import confusion_matrix
+predicted_classes = np.argmax(model.predict(X_valid_subset), axis=-1)
+y_true = np.argmax(y_valid_subset, axis=-1)
+
+cm = confusion_matrix(y_true, predicted_classes)
+
+plt.figure(figsize = (25, 25))
+sns.heatmap(cm, annot = True)
+
+plt.show()

TestOnePictureValid.py

@@ -7,21 +7,18 @@ import seaborn as sns
 import pickle
 import random
 import os
+import numpy as np
+from keras.utils import np_utils
 
-with open(os.path.join("dataset", "train.p"), mode='rb') as training_data:
-    train = pickle.load(training_data)
 with open(os.path.join("dataset", "valid.p"), mode='rb') as validation_data:
     valid = pickle.load(validation_data)
 
-X_train, y_train = train['features'], train['labels']
 X_valid, y_valid = valid['features'], valid['labels']
 
 from sklearn.utils import shuffle
-X_train, y_train = shuffle(X_train, y_train)
 X_valid, y_valid = shuffle(X_valid, y_valid)
 
 # Normalize image to [0, 1]
-X_train_norm = X_train / 255
 X_valid_norm = X_valid / 255
 
 #TakeOnePicture (Klasse 0-19)
@@ -29,7 +26,10 @@ filtered_indices = [i for i, label in enumerate(y_valid) if label >= 0 and label
 indice = random.choice(filtered_indices)
 X_valid_subset = X_valid[indice][np.newaxis, ...]
 y_valid_subset = y_valid[indice][np.newaxis, ...]
-print(y_valid_subset)
+
+num_classes = 20  # Anzahl der Straßenschilder-Klassen
+y_valid_subset = np_utils.to_categorical(y_valid_subset, num_classes)
+
 
 
 #ownModel
@@ -44,7 +44,7 @@ print('Test Accuracy: {}'.format(score[1]))
 
 from sklearn.metrics import confusion_matrix
 predicted_classes = np.argmax(model.predict(X_valid_subset), axis=-1)
-y_true = y_valid_subset
+y_true = np.argmax(y_valid_subset, axis=-1)
 
 cm = confusion_matrix(y_true, predicted_classes)