Edited model, need to change load_images.py to expand waldo data
This commit is contained in:
Silver-T
parent
b6892721e1
commit
ec823fe688
@ -19,6 +19,7 @@ def gen_data(w_path, n_w_path):
|
||||
i = 0
|
||||
for image_name in waldo_file_list:
|
||||
pic = cv2.imread(os.path.join(w_path, image_name)) # NOTE: cv2.imread() returns a numpy array in BGR not RGB
|
||||
pic = pic/255 # Scaling images down to values of 0-255
|
||||
imgs_raw.append(np.rollaxis(pic, -1)) # rolls colour axis to 0
|
||||
imgs_lbl.append(1) # Value of 1 as Waldo is present in the image
|
||||
|
||||
@ -28,22 +29,48 @@ def gen_data(w_path, n_w_path):
|
||||
i = 0
|
||||
for image_name in not_waldo_file_list:
|
||||
pic = cv2.imread(os.path.join(n_w_path, image_name))
|
||||
pic = pic/255 # Scaling images down to values of 0-255
|
||||
imgs_raw.append(np.rollaxis(pic, -1))
|
||||
imgs_lbl.append(0)
|
||||
|
||||
print('Completed: {0}/{1} non-Waldo images'.format(i+1, total_nw))
|
||||
i += 1
|
||||
|
||||
# Calculate what 30% of each set is
|
||||
third_of_w = math.floor(0.3*total_w)
|
||||
third_of_nw = math.floor(0.3*total_nw)
|
||||
## Randomise and split data into training and test sets
|
||||
# Code was modified from code written by: Kyle O'Brien (medium.com/@kylepob61392)
|
||||
n_images = len(imgs_raw)
|
||||
TRAIN_TEST_SPLIT = 0.75
|
||||
|
||||
# Split at the given index
|
||||
split_index = int(TRAIN_TEST_SPLIT * n_images)
|
||||
shuffled_indices = np.random.permutation(n_images)
|
||||
train_indices = shuffled_indices[0:split_index]
|
||||
test_indices = shuffled_indices[split_index:]
|
||||
|
||||
train_data = []
|
||||
train_lbl = []
|
||||
test_data = []
|
||||
test_lbl = []
|
||||
|
||||
# Split the images and the labels
|
||||
for index in train_indices:
|
||||
train_data.append(imgs_raw[index])
|
||||
train_lbl.append(imgs_lbl[index])
|
||||
|
||||
for index in test_indices:
|
||||
test_data.append(imgs_raw[index])
|
||||
test_lbl.append(imgs_lbl[index])
|
||||
|
||||
# # Calculate what 30% of each set is
|
||||
# third_of_w = math.floor(0.3*total_w)
|
||||
# third_of_nw = math.floor(0.3*total_nw)
|
||||
|
||||
# Split data into training and test data (60%/30%)
|
||||
train_data = np.append(imgs_raw[(third_of_w+1):total_w], imgs_raw[(total_w + third_of_nw + 1):len(imgs_raw)-1], axis=0)
|
||||
train_lbl = np.append(imgs_lbl[(third_of_w+1):total_w], imgs_lbl[(total_w + third_of_nw + 1):len(imgs_lbl)-1], axis=0)
|
||||
# If axis not given, both arrays are flattened before being appended
|
||||
test_data = np.append(imgs_raw[0:third_of_w], imgs_raw[total_w:(total_w + third_of_nw)], axis=0)
|
||||
test_lbl = np.append(imgs_lbl[0:third_of_w], imgs_lbl[total_w:(total_w + third_of_nw)], axis=0)
|
||||
# # Split data into training and test data (60%/30%)
|
||||
# train_data = np.append(imgs_raw[(third_of_w+1):total_w], imgs_raw[(total_w + third_of_nw + 1):len(imgs_raw)-1], axis=0)
|
||||
# train_lbl = np.append(imgs_lbl[(third_of_w+1):total_w], imgs_lbl[(total_w + third_of_nw + 1):len(imgs_lbl)-1], axis=0)
|
||||
# # If axis not given, both arrays are flattened before being appended
|
||||
# test_data = np.append(imgs_raw[0:third_of_w], imgs_raw[total_w:(total_w + third_of_nw)], axis=0)
|
||||
# test_lbl = np.append(imgs_lbl[0:third_of_w], imgs_lbl[total_w:(total_w + third_of_nw)], axis=0)
|
||||
|
||||
try:
|
||||
# Save the data as numpy files
|
||||
|
||||
@ -10,14 +10,16 @@ from keras.layers import Conv2D, MaxPooling2D, ZeroPadding2D
|
||||
from keras.models import Model
|
||||
|
||||
from sklearn import svm, tree, naive_bayes, ensemble
|
||||
from sklearn.metrics import accuracy_score
|
||||
from _image_classifier import ImageClassifier
|
||||
|
||||
from keras.optimizers import Adadelta
|
||||
from keras.optimizers import Adam
|
||||
from keras.callbacks import ModelCheckpoint
|
||||
from keras import backend as K
|
||||
K.set_image_dim_ordering('th')
|
||||
np.random.seed(7)
|
||||
|
||||
from keras.utils import to_categorical
|
||||
'''
|
||||
Model definition define the network structure
|
||||
'''
|
||||
@ -25,28 +27,28 @@ def FCN():
|
||||
## List of model layers
|
||||
inputs = Input((3, 64, 64))
|
||||
|
||||
conv1 = Conv2D(8, (2, 2), activation='relu', padding='same', input_shape=(64, 64, 3))(inputs)
|
||||
conv1 = Conv2D(16, (3, 3), activation='relu', padding='same', input_shape=(64, 64, 3))(inputs)
|
||||
m_pool1 = MaxPooling2D(pool_size=(2, 2))(conv1)
|
||||
|
||||
conv2 = Conv2D(16, (2, 2), activation='relu', padding='same')(m_pool1)
|
||||
drop1 = Dropout(0.2)(conv2) # Drop some portion of features to prevent overfitting
|
||||
m_pool2 = MaxPooling2D(pool_size=(2, 2))(drop1)
|
||||
conv2 = Conv2D(32, (3, 3), activation='relu', padding='same')(m_pool1)
|
||||
#drop1 = Dropout(0.2)(conv2) # Drop some portion of features to prevent overfitting
|
||||
m_pool2 = MaxPooling2D(pool_size=(2, 2))(conv2)
|
||||
|
||||
conv3 = Conv2D(32, (2, 2), activation='relu', padding='same')(m_pool2)
|
||||
drop2 = Dropout(0.2)(conv3) # Drop some portion of features to prevent overfitting
|
||||
m_pool2 = MaxPooling2D(pool_size=(2, 2))(drop2)
|
||||
conv3 = Conv2D(32, (3, 3), activation='relu', padding='same')(m_pool2)
|
||||
# drop2 = Dropout(0.2)(conv3) # Drop some portion of features to prevent overfitting
|
||||
# m_pool2 = MaxPooling2D(pool_size=(2, 2))(drop2)
|
||||
|
||||
conv4 = Conv2D(164, (2, 2), activation='relu', padding='same')(m_pool2)
|
||||
# conv4 = Conv2D(64, (2, 2), activation='relu', padding='same')(m_pool2)
|
||||
|
||||
flat = Flatten()(conv4) # Makes data 1D
|
||||
flat = Flatten()(conv3) # Makes data 1D
|
||||
dense = Dense(64, activation='relu')(flat) # Fully connected layer
|
||||
drop3 = Dropout(0.2)(dense)
|
||||
classif = Dense(2, activation='softmax')(drop3) # Final layer to classify
|
||||
classif = Dense(2, activation='sigmoid')(drop3) # Final layer to classify
|
||||
|
||||
## Define the model structure
|
||||
model = Model(inputs=inputs, outputs=classif)
|
||||
# Optimizer recommended Adadelta values (lr=0.01)
|
||||
model.compile(optimizer=Adadelta(lr=0.1), loss='sparse_categorical_crossentropy', metrics=['accuracy'])
|
||||
model.compile(optimizer=Adam(), loss='binary_crossentropy', metrics=['accuracy'])
|
||||
|
||||
return model
|
||||
|
||||
@ -57,37 +59,36 @@ lbl_train = np.load('Waldo_train_lbl.npy')
|
||||
im_test = np.load('Waldo_test_data.npy')
|
||||
lbl_test = np.load('Waldo_test_lbl.npy')
|
||||
|
||||
lbl_train = to_categorical(lbl_train) # One hot encoding the labels
|
||||
lbl_test = to_categorical(lbl_test)
|
||||
|
||||
## Define model
|
||||
model = FCN()
|
||||
svm_iclf = ImageClassifier(svm.SVC)
|
||||
tree_iclf = ImageClassifier(tree.DecisionTreeClassifier)
|
||||
naive_bayes_iclf = ImageClassifier(naive_bayes.GaussianNBd)
|
||||
ensemble_iclf = ImageClassifier(ensemble.RandomForestClassifier)
|
||||
# svm_iclf = ImageClassifier(svm.SVC)
|
||||
# tree_iclf = ImageClassifier(tree.DecisionTreeClassifier)
|
||||
# naive_bayes_iclf = ImageClassifier(naive_bayes.GaussianNBd)
|
||||
# ensemble_iclf = ImageClassifier(ensemble.RandomForestClassifier)
|
||||
|
||||
## Define training parameters
|
||||
epochs = 20 # an epoch is one forward pass and back propogation of all training data
|
||||
batch_size = 5
|
||||
#lrate = 0.01
|
||||
#decay = lrate/epochs
|
||||
# epoch - one forward pass and one backward pass of all training data
|
||||
# batch size - number of training example used in one forward/backward pass
|
||||
# (higher batch size uses more memory)
|
||||
# learning rate - controls magnitude of weight changes in training the NN
|
||||
batch_size = 150 # batch size - number of training example used in one forward/backward pass
|
||||
# (higher batch size uses more memory, smaller batch size takes more time)
|
||||
#lrate = 0.01 # Learning rate of the model - controls magnitude of weight changes in training the NN
|
||||
#decay = lrate/epochs # Decay rate of the model
|
||||
|
||||
## Train model
|
||||
# Purely superficial output
|
||||
sys.stdout.write("\nFitting model")
|
||||
sys.stdout.flush()
|
||||
for i in range(0, 3):
|
||||
t.sleep(0.8)
|
||||
t.sleep(0.5)
|
||||
sys.stdout.write('.')
|
||||
sys.stdout.flush()
|
||||
t.sleep(0.5)
|
||||
print()
|
||||
|
||||
# Outputs the model structure
|
||||
for i in range(0, len(model.layers)):
|
||||
print("Layer {}: {}".format(i, model.layers[i].output))
|
||||
print('-'*30)
|
||||
print(model.summary())
|
||||
|
||||
filepath = "checkpoint.hdf5" # Defines the model checkpoint file
|
||||
checkpoint = ModelCheckpoint(filepath, verbose=1, save_best_only=False) # Defines the checkpoint process
|
||||
@ -120,15 +121,13 @@ model.save('Waldo.h5')
|
||||
print("\nModel weights and structure have been saved.\n")
|
||||
|
||||
## Testing the model
|
||||
# Show data stats
|
||||
print('*'*30)
|
||||
print(im_test.shape)
|
||||
print(lbl_test.shape)
|
||||
print('*'*30)
|
||||
start = t.time()
|
||||
# Passes the dataset through the model
|
||||
pred_lbl = model.predict(im_test, verbose=1, batch_size=batch_size)
|
||||
end = t.time()
|
||||
pred_lbl = np.round(pred_lbl)
|
||||
accuracy = accuracy_score(lbl_test, pred_lbl)
|
||||
print("Accuracy: " + str(accuracy))
|
||||
print("Images generated in {} seconds".format(end - start))
|
||||
np.save('predicted_results.npy', pred_lbl)
|
||||
|
||||
|
||||
Reference in New Issue
Block a user