def testDataloader():
from src import dataloader
import cv2
data, labels, classes = dataloader.loadData("data_part1/train/")
td, tl, vd, vl = dataloader.splitValidation(data, labels, 10)
print(tl[0])
cv2.imshow('Imagem', td[0])
cv2.waitKey(0)
cv2.destroyAllWindows()
def validate(x, y, w0, b0):
ok = 0
for i in range(len(x)):
y_ = sigmoid(np.dot(x[i], w0) + b0)
shot = np.argmax(y_)
if (y[i][shot] == 1):
ok += 1
return ok / len(x)
if __name__ == '__main__': # main here
data, labels, classes = dataloader.loadData("data_part1/train/")
td, tl, vd, vl = dataloader.splitValidation(data, labels, 10)
# normalizing train and validation data [0, 1]
td = np.reshape(td, (len(td), 77 * 71))
vd = np.reshape(vd, (len(vd), 77 * 71))
td = td / 255.
vd = vd / 255.
epoch = 130
learning_rate = 0.1
batch_size = 20
# 10 dimensoes para os pesos
w = np.random.uniform(-0.1, 0.1, (len(td[0]), 10))
b = np.zeros(10)
#print(w[0])
IMAGE_HEIGHT = 64 # height of the image
IMAGE_WIDTH = 64 # width of the image
NUM_CHANNELS = 1 # number of channels of the image
NUM_EPOCHS_FULL = 200
S_LEARNING_RATE_FULL = 0.001
F_LEARNING_RATE_FULL = 0.001
BATCH_SIZE = 64
data, labels, classes = dataloader.loadData(TRAIN_PATH)
resized = dataloader.resize(data, IMAGE_WIDTH, IMAGE_HEIGHT)
# cv2.imshow('resized', resized[10])
# cv2.waitKey(0)
# exit()
td, tl, vd, vl = dataloader.splitValidation(resized, labels, 10)
td = np.reshape(td, (len(td), IMAGE_HEIGHT, IMAGE_WIDTH, NUM_CHANNELS))/255
vd = np.reshape(vd, (len(vd), IMAGE_HEIGHT, IMAGE_WIDTH, NUM_CHANNELS))/255
discClasses = np.array(['fake', 'true'])
graph = tf.Graph()
with graph.as_default():
# image input
X = tf.placeholder(tf.float32, shape = (None, IMAGE_HEIGHT, IMAGE_WIDTH, NUM_CHANNELS))
# discrimator y
y = tf.placeholder(tf.float32, shape = (None))