def post():
images = get_images.main()
while images is None:
images = get_images.main()
impath, source, text = images
with open(impath, "rb") as img:
response = twitter.upload_media(media = img)
message = f"{text} ({source})"
out = twitter.update_status(status=message, media_ids=[response["media_id"]])
get_images.logging.info("Posted to twitter.")
def download(self, event):
url = self.url.GetValue()
dir_ = self.dir.GetValue()
get_images.main(url, dir_)
from pages import git_pull_origin import get_gists_from_user as get_gists import get_images git_pull_origin.main() get_gists.main() get_images.main()
from pages import git_pull_origin import get_gists_from_user as get_gists import get_images git_pull_origin.main() get_gists.main() get_images.main()
def download(self, event):
url = self.url.GetValue()
dir_ = self.dir.GetValue()
get_images.main(url, dir_)
def main(mode, n_classes, im_size, prj_dir, feat_dir, classes, batch_size=50):
print('Getting data...')
train_datagen = tf.keras.preprocessing.image.ImageDataGenerator(
featurewise_center=True, featurewise_std_normalization=True)
valid_datagen = tf.keras.preprocessing.image.ImageDataGenerator(
featurewise_center=True, featurewise_std_normalization=True)
test_datagen = tf.keras.preprocessing.image.ImageDataGenerator(
featurewise_center=True, featurewise_std_normalization=True)
images = get_images.main(feat_dir + r'\\train')
train_datagen.fit(images)
valid_datagen.fit(images)
test_datagen.fit(images)
del images
train_batches = train_datagen.flow_from_directory(feat_dir + r'\\train',
target_size=im_size,
batch_size=batch_size,
classes=classes)
valid_batches = valid_datagen.flow_from_directory(feat_dir + r'\\valid',
target_size=im_size,
batch_size=batch_size,
classes=classes)
test_batches = test_datagen.flow_from_directory(feat_dir + r'\\test',
target_size=im_size,
batch_size=batch_size,
classes=classes,
shuffle=False)
true_test = test_batches.classes
test_names = test_batches.filenames
# create log folder if not exists:
if not os.path.exists(prj_dir + r'\\logs\\'):
os.makedirs(prj_dir + r'\\logs\\')
# create model folder if not exists:
if not os.path.exists(prj_dir + r'\\model\\'):
os.makedirs(prj_dir + r'\\model\\')
log_dir = prj_dir + r"\\logs\\" + datetime.datetime.now().strftime(
"%Y%m%d-%H%M%S")
checkpoint_dir = prj_dir + r".\\model\\model.{epoch:02d}-{val_loss:.2f}.h5"
input_shape = (im_size[0], im_size[1], 3)
if mode is "train":
print('Start training...')
model = train(log_dir, checkpoint_dir, n_classes, train_batches,
valid_batches, input_shape)
print('Done training!')
print('Start Testing...')
pred = model.predict(x=test_batches, verbose=0)
print('Done testing!')
if mode is 'test':
# get saved model:
print('Import model...')
list_of_files = glob.glob(
prj_dir +
r'\model\*') # * means all if need specific format then *.csv
latest = max(list_of_files, key=os.path.getctime)
# Create a new model instance
model = create_model(n_classes, input_shape)
# Load the previously saved weights
model.load_weights(latest)
print('Start Testing...')
pred = model.predict(x=test_batches, verbose=0)
print('Done testing!')
return pred, true_test, test_names
def get_images_(args):
import get_images
get_images.main(args.owner_id)
