def recognize_v2(target):
model = create_model()
#loading weights
if os.path.exists('weights2.h5'):
model.load_weights('weights2.h5')
else:
train(model)
if os.path.isfile(target):
img = image.load_img(target, target_size=(img_width, img_height))
x = image.img_to_array(img)
x = np.expand_dims(x, axis=0)
prediction = model.predict(x)
raw_img = matplotlib.image.imread(target)
plt.imshow(raw_img)
if prediction:
result = 'cat'
plt.text(0, -20, 'Я не хлеб, я кот! Не ешьте меня!', fontsize=20)
else:
result = 'bread'
plt.text(50, -60, 'Это сладкий хлебушек.', fontsize=20)
plt.text(50, -5, 'Кушайте на здоровье!', fontsize=20)
plt.axis("off")
plt.show()
else:
raise IOError('No such file')
def recognize(target):
model = create_model()
#loading weights
if os.path.exists('weights.h5'):
model.load_weights('weights.h5')
else:
train(model)
if os.path.isfile(target):
img = image.load_img(target, target_size=(img_width, img_height))
x = image.img_to_array(img)
x = np.expand_dims(x, axis=0)
prediction = model.predict(x)
raw_img = matplotlib.image.imread(target)
plt.imshow(raw_img)
if prediction:
result = 'butterflies'
plt.text(0, -20, 'Это изображение бабочки.', fontsize=20)
else:
result = 'flowers'
plt.text(150, -50, 'Это изображение цветов.', fontsize=20)
plt.axis("off")
plt.show()
else:
raise IOError('No such file')
def predict_traffic(self):
print('Predict traffic')
label = self.picture_dic[self.current_pic]
if self.mode == 'labeled_images':
pic_path = os.path.join(pic_dir, label)
pic_path = os.path.join(pic_path, self.current_pic)
else:
pic_path = os.path.join(pic_dir, self.current_pic)
pic = image.load_img(pic_path, target_size=(150, 150))
pic_array = image.img_to_array(pic)
pic_array = pic_array / 255
img = np.expand_dims(pic_array, axis=0)
print(img.shape)
result = model.predict_classes(img)
prediction = result[0]
if prediction == 0:
prediction = 'low'
else:
prediction = 'medium'
print(prediction)
if prediction == label:
self.correct_preds = self.correct_preds + 1
self.total_preds = self.total_preds + 1
self.prediction_label['text'] = 'Traffic Prediction: {}'.format(
prediction)
self.answer_label['text'] = 'Traffic Answer: ' + label
self.acc_label['text'] = 'Acc: ' + str(
(self.correct_preds / self.total_preds * 100)) + '%'
def run_demo():
model = create_model()
# loading weights
if os.path.exists('weights.h5'):
model.load_weights('weights.h5')
else:
train(model)
# validation and test augmentation. Only rescaling
test_datagen = ImageDataGenerator(rescale=1. / 255)
pred_generator = test_datagen.flow_from_directory(validation_data_dir,
target_size=(150, 150),
batch_size=100,
class_mode='binary')
imgs, labels = pred_generator.next()
array_imgs = np.asarray([image.img_to_array(img) for img in imgs])
predictions = model.predict(imgs)
rounded_pred = np.asarray([np.round(i) for i in predictions])
result = [im for im in zip(array_imgs, rounded_pred, labels, predictions)]
wrong = [x for x in result if x[1] != x[2]]
mistake = len(wrong) / len(result)
accuracy = 100 - mistake * 100
print(len(wrong))
print(len(result))
print('Mistake -- {}%'.format(mistake * 100))
plt.figure(figsize=(12, 12))
plt.figtext(0,
0,
' Точность -- {}%'.format(accuracy),
fontsize=20)
for ind, val in enumerate(result[:16]):
plt.subplot(4, 4, ind + 1)
im = val[0]
if int(val[1]):
lb = 'Кот'
cl = 'blue'
else:
lb = 'Хлеб'
cl = 'red'
plt.axis('off')
plt.text(50, -4, lb, fontsize=20, color=cl)
plt.imshow(np.transpose(im, (0, 1, 2)))
plt.show()
def recognize(target):
model = create_model()
#loading weights
model.load_weights(
'/Users/alexivannikov/recognizeMelanoma/recognizeApp/networks/weights_new.h5'
)
if os.path.isfile(target):
img = image.load_img(target, target_size=(256, 256))
x = image.img_to_array(img)
x = x[np.newaxis, ...]
img2 = imread(target)
prediction = model.predict(x)
print(int(prediction))
print(prediction[0][0])
K.clear_session()
return prediction[0][0]
def run_demo_v2():
model = create_model()
# loading weights
if os.path.exists('weights2.h5'):
model.load_weights('weights2.h5')
else:
train(model)
# validation and test augmentation. Only rescaling
test_datagen = ImageDataGenerator(rescale=1. / 255)
pred_generator = test_datagen.flow_from_directory(validation_data_dir,
target_size=(150, 150),
batch_size=100,
class_mode='binary')
imgs, labels = pred_generator.next()
array_imgs = np.asarray([image.img_to_array(img) for img in imgs])
predictions = model.predict(imgs)
rounded_pred = np.asarray([np.round(i) for i in predictions])
result = [im for im in zip(array_imgs, rounded_pred, labels, predictions)]
for ind, val in enumerate(result[:16]):
plt.subplot(4, 4, ind + 1)
im = val[0]
if int(val[1]):
lb = 'Кот'
cl = 'blue'
else:
lb = 'Хлеб'
cl = 'red'
plt.axis('off')
plt.text(50, -4, lb, fontsize=20, color=cl)
plt.imshow(np.transpose(im, (0, 1, 2)))
plt.show()
autoencoder.fit(x_train, x_train,
epochs=500,
batch_size=25,
shuffle=True,
validation_data=(x_train, x_train),)
decoded_imgs = autoencoder.predict(x_test)
## Check Image
# create a data generator
datagen = ImageDataGenerator()
train_it = datagen.flow_from_directory('.',classes=['Images Night'],target_size=(28, 28))
img_path = 'Images Night'
img = list()
for filename in listdir('Images Night'):
img_tensor=image.load_img('Images Night\\' + filename, target_size=(28, 28))
img_tensor = image.img_to_array(img_tensor)
img_tensor = np.expand_dims(img_tensor, axis=0)
img_tensor /= 255.
img.append(img_tensor)
plt.imshow(img_tensor[0])
plt.show()
print(img_tensor.shape)
## check intermediat result
layer_outputs = [layer.output for layer in autoencoder.layers[:12]] # Extracts the outputs of the top 12 layers
activation_model = models.Model(inputs=autoencoder.input, outputs=layer_outputs) # Creates a model that will return these outputs, given the model input
activations = activation_model.predict(img_tensor)
first_layer_activation = activations[0]
print(first_layer_activation.shape)