def main():
print('Running main function')
# Loading images in color
width_resize = 28
height_resize = 28
instance_load = ClassImageDataSet(positive_dir, negative_dir, width_resize,
height_resize)
train_data, train_labels = instance_load.load_train_set(load_color=True)
eval_data, eval_labels = instance_load.load_eval_set(load_color=True)
print('Printing shapes')
print(train_data.shape)
print(train_labels.shape)
print('Generating instance')
model_dir = '/tmp/cnn_color'
n_classes = 2
channels = 3
classifier = ClassCNN(model_dir=model_dir,
classes=n_classes,
width=width_resize,
height=height_resize,
channels=channels,
train_steps=5000)
var = input('Set 1 to train, 2 to predict. Otherwise to eval ')
if var == '1':
print('Training model')
classifier.train_model(train_data, train_labels)
elif var == '2':
print('Predict model')
print('Total elements: ' + str(eval_data.shape[0]))
index = 0
eval_item = eval_data[index]
print(eval_item.shape)
result = classifier.predict_model(eval_item)
print('Result obtained: ' + str(result['classes']))
print('Print probabilities')
print(result['probabilities'])
print('Real result: ' + str(eval_labels[index]))
else:
print('Evaluating model')
classifier.eval_model(eval_data, eval_labels)
print('Done!')
def train_model_one_vs_all():
print('Training MNIST model one vs all')
train_data, train_labels = ClassImageDataSet.load_train_mnist()
data = train_data[0]
print('Data: {0}'.format(data))
classes = 10
list_models = list()
for cls in range(classes):
list_labels = list()
for i in range(train_labels.shape[0]):
if cls == int(train_labels[i]):
list_labels.append(1)
else:
list_labels.append(0)
new_labels_np = np.array(list_labels, dtype=np.int)
print('Generating values to train in SVM: {0}'.format(cls))
start_time = time.time()
clf = svm.SVC(gamma=0.001, C=100.)
clf.fit(train_data, new_labels_np)
end_time = time.time()
print('Total time: {0}'.format((end_time - start_time)))
path_model = '/tmp/model{0}.pkl'.format(cls)
print('Generating model into: {0}'.format(path_model))
joblib.dump(clf, path_model)
list_models.append(clf)
print('Evaluating model list')
eval_data, eval_labels = ClassImageDataSet.load_eval_mnist()
print('Getting first data!')
print('Label: {0}'.format(eval_labels[0]))
for index, model in enumerate(list_models):
array = np.expand_dims(eval_data[0], axis=0)
result = model.predict(array)
print('Result for class {0}: {1}'.format(index, result))
print('Done!')
def main():
print('Loading dataset')
width_resize = 28
height_resize = 28
instance = ClassImageDataSet(positive_dir, negative_dir, width_resize, height_resize)
pos_train, neg_train = instance.load_train_set()
pos_eval, neg_eval = instance.load_eval_set()
print('Test loading mnist dataset')
pos_mnist_train, pos_mnist_eval = instance.load_eval_mnist()
print('Train dataset')
print(pos_mnist_train[0])
print('Done!')
def eval_model():
print('Evaluating MNIST model')
print('Loading model')
clf = joblib.load('/tmp/model.pkl') # type: svm.SVC
print('Loading train data')
test, labels_test = ClassImageDataSet.load_eval_mnist()
elem = test[0]
print('Evaluating model in elements')
elems = 0
pos = 0
for elem in test:
array = np.expand_dims(elem, axis=0)
prediction = clf.predict(array)
if prediction[0] == labels_test[elems]:
pos = pos + 1
elems = elems + 1
if elems % 10 == 0:
print('Elems: ' + str(elems) + ' pos: ' + str(pos) + ' total: ' +
str(labels_test.shape))
precision = pos / elems
print('Pos: ' + str(pos))
print('Elems: ' + str(elems))
print('Precision: ' + str(precision))
print('Done!')
def main():
print('Init main function')
print('Loading models')
print('Load model in grayscale first')
width = 28
height = 28
# Test as reshape image
train_data, train_labels = ClassImageDataSet.load_train_mnist(reshape=True)
eval_data, eval_labels = ClassImageDataSet.load_eval_mnist(reshape=True)
n_classes = 10
channels = 1
model_dir = '/tmp/model_example_mnist'
classifier = ClassCNN(model_dir=model_dir,
channels=channels,
classes=n_classes,
width=width,
height=height,
train_steps=5000)
var = input('Set 1 to train, 2 to predict. Otherwise to eval ')
if var == '1':
print('Training model')
classifier.train_model(train_data, train_labels)
elif var == '2':
print('Predict model')
print('Total elements: ' + str(eval_data.shape[0]))
index = 1100
eval_item = eval_data[index]
print(eval_item.shape)
result = classifier.predict_model(eval_item)
print('Result obtained: ' + str(result['classes']))
print('Print probabilities')
print(result['probabilities'])
print('Real result: ' + str(eval_labels[index]))
else:
print('Evaluating model')
classifier.eval_model(eval_data, eval_labels)
print('Done!')
def main():
print('Initializing main function')
print('Loading datasets')
train_data, train_labels = ClassImageDataSet.load_train_mnist()
eval_data, eval_labels = ClassImageDataSet.load_eval_mnist()
print('PCA with training data')
n_features = 18
pca = PCA(n_components=n_features, svd_solver='randomized').fit(train_data)
train_pca = pca.transform(train_data)
n_classes = 10
hidden_neurons = 100
eval_pca = pca.transform(eval_data)
print('Printing shapes')
print(train_data.shape)
print(train_pca.shape)
model_dir = '/tmp/model_example_pca'
classifier = ClassNN(model_dir, n_classes, hidden_neurons)
var = input('Set 1 to train, 2 to predict. Otherwise to eval ')
if var == '1':
print('Training model')
classifier.train_model(train_pca, train_labels)
elif var == '2':
print('Predict model')
print('Total elements: ' + str(eval_pca.shape[0]))
index = 1100
eval_item = eval_pca[index]
print(eval_item.shape)
result = classifier.predict_model(eval_item)
print('Result obtained: ' + str(result['classes']))
print('Print probabilities')
print(result['probabilities'])
print('Real result: ' + str(eval_labels[index]))
else:
print('Evaluating model')
classifier.eval_model(eval_pca, eval_labels)
print('Done!')
def main():
print('Initializing main function')
instance_train = ClassImageDataSet(positive_dir, negative_dir, width_resize, height_resize)
train_data, train_labels = instance_train.load_train_mnist()
eval_data, eval_labels = instance_train.load_eval_mnist()
res = input('Press 1 to train, 2 to eval, 3 to eval fast, 4 to eval fast pose: ')
if res == '1':
train(train_data, train_labels)
elif res == '2':
evaluating(eval_data)
elif res == '3':
evaluating_fast(eval_data)
elif res == '4':
evaluating_fast_nn()
else:
raise Exception('Option not recognized!')
def train_model():
print('Training MNIST model')
train, labels = ClassImageDataSet.load_train_mnist()
print('Generating values to train in SVM')
clf = svm.SVC(gamma=0.001, C=100.)
clf.fit(train, labels)
print('Saving model in temp folder')
joblib.dump(clf, '/tmp/model.pkl')
def main():
print('Initializing main function')
train_data, train_labels = ClassImageDataSet.load_train_mnist()
components = 48
pca = PCA(n_components=components,
svd_solver='randomized').fit(train_data) # type: PCA
train_pca = pca.transform(train_data)
print(pca.explained_variance_ratio_.sum())
plt.hist(pca.explained_variance_ratio_, bins=components, log=True)
plt.show()
print('Done!')
def eval_model_one_vs_all():
print('Evaluating SVN multi class')
print('Loading models')
classes = 10
list_models = list()
for cls in range(classes):
path_model = '/tmp/model{0}.pkl'.format(cls)
print('Loading model {0}'.format(cls))
model = joblib.load(path_model)
list_models.append(model)
# Loading eval classes
eval_data, eval_labels = ClassImageDataSet.load_eval_mnist()
data_ok = 0
total_data = eval_labels.shape[0]
for index in range(total_data):
data = eval_data[index]
array = np.expand_dims(data, axis=0)
selected_cls = -1
for cls in range(classes):
model = list_models[cls]
res = model.predict(array)
if res[0] == 1:
selected_cls = cls
if selected_cls == int(eval_labels[index]):
data_ok += 1
print('Data ok: {0}'.format(data_ok))
print('Total data: {0}'.format(total_data))
precision = data_ok / total_data
print('Precision: {0}'.format(precision))
print('Done!')