def recognize(image_path: str,
weights_path: str,
config: GlobalConfig,
is_vis=True):
logger = LogFactory.get_logger()
image = load_and_resize_image(image_path)
inputdata = tf.placeholder(dtype=tf.float32,
shape=[1, 32, 100, 3],
name='input')
net = CRNN(phase='Test', hidden_nums=256, seq_length=25, num_classes=37)
with tf.variable_scope('shadow'):
net_out = net.build(inputdata=inputdata)
decodes, _ = tf.nn.ctc_beam_search_decoder(inputs=net_out,
sequence_length=25 * np.ones(1),
merge_repeated=False)
decoder = TextFeatureIO()
# config tf session
sess_config = tf.ConfigProto()
sess_config.gpu_options.per_process_gpu_memory_fraction = config.get_gpu_config(
).memory_fraction
sess_config.gpu_options.allow_growth = config.get_gpu_config(
).is_tf_growth_allowed()
# config tf saver
saver = tf.train.Saver()
sess = tf.Session(config=sess_config)
with sess.as_default():
saver.restore(sess=sess, save_path=weights_path)
preds = sess.run(decodes, feed_dict={inputdata: image})
preds = decoder.writer.sparse_tensor_to_str(preds[0])
logger.info('Predict image {:s} label {:s}'.format(
ops.split(image_path)[1], preds[0]))
if is_vis:
plt.figure('CRNN Model Demo')
plt.imshow(
cv2.imread(image_path, cv2.IMREAD_COLOR)[:, :, (2, 1, 0)])
plt.show()
sess.close()
def __init__(self, screen):
# Call the super constructor
pygame.sprite.Sprite.__init__(self)
self.screen = screen
# Load and resize the star image
self.image = utils.load_and_resize_image(
"resources/star.png", (Star.STAR_SIZE, Star.STAR_SIZE))
self.image.fill(
self.COLOR, None,
pygame.BLEND_MIN) # Fill the star image with the score color
# Set initial score at 0
self.score = 0
# Open the font for the score
self.font = pygame.font.Font("resources/Bonk.ttf", 50)
def switch_ball_color(self):
image_file_path = utils.random_file_from_folder("resources/ball")
# While the new image file is the same as the previous image, request a new one
while image_file_path == self.image_file:
# Get a random ball image from the ball images folder
image_file_path = utils.random_file_from_folder("resources/ball")
# Set the image file as the new image file path
self.image_file = image_file_path
# Load the ball image and resize it to it's size by the diameter + border
self.image = utils.load_and_resize_image(
self.image_file, (self.BALL_DIAMETER + self.BALL_BORDER,
self.BALL_DIAMETER + self.BALL_BORDER))
# Get the most used color in the image and use it as the ball's color
self.color = utils.get_dominant_color_from_image(self.image_file)
def __init__(self, screen, image, size, y_pos):
# Call the super constructor
pygame.sprite.Sprite.__init__(self)
# Save the screen as a field
self.screen = screen
# Set the height and the width by the sprite's size
self.width = size[0]
self.height = size[1]
# Create the sprite's rect and set the sprite in the middle of the screen
self.rect = pygame.Rect(screen.get_rect().width / 2 - self.width / 2, y_pos, self.width, self.height)
# If an image was sent, set it
if image is not None:
# Load the image and resize it to the correct size
self.image = utils.load_and_resize_image(image, (self.width, self.height))
from utils import load_and_resize_image, get_data, k_fold_cross_val, train_test_split_sk
from sklearn.metrics import accuracy_score
TRAIN_DIR = "images/train"
TEST_DIR = "images/test"
INPUT_SHAPE = (100, 100, 3)
# Reading the image
train_names, train_val, train_paths = get_data(TRAIN_DIR)
test_names, test_val, test_paths = get_data(TEST_DIR)
train_images = []
test_images = []
for img in train_paths:
train_images.append(load_and_resize_image(img, INPUT_SHAPE))
for img in test_paths:
test_images.append(load_and_resize_image(img, INPUT_SHAPE))
train_images = np.array(train_images)
test_images = np.array(test_images)
train_val = np.array(train_val)
test_val = np.array(test_val)
# Define model
model = CNNClass.Sequential(input_shape=INPUT_SHAPE)
# Convoluting the image
print("**Convolution Layer Start**")
print("**Convolution Stage**")
def parse_params():
parser = argparse.ArgumentParser()
parser.add_argument('-i', '--image_path', type=str, metavar='PATH')
parser.add_argument('-m', '--model', type=str, metavar='PATH')
return parser.parse_args()
if __name__ == '__main__':
params = parse_params()
model_file = params.model
image_path = params.image_path
if not exists(model_file):
raise ValueError('{:s} doesn\'t exist'.format(model_file))
image = load_and_resize_image(image_path)
padded_images = np.zeros([BATCH_SIZE, 32, 100, 3])
padded_images[:image.shape[0], :, :, :] = image
tf.reset_default_graph()
with tf.Session(graph=tf.Graph()) as sess:
with gfile.FastGFile(model_file, 'rb') as f:
graph_def = tf.GraphDef()
graph_def.ParseFromString(f.read())
sess.graph.as_default()
tf.import_graph_def(graph_def, name='')
inputdata = tf.get_default_graph().get_tensor_by_name("input:0")
output = tf.get_default_graph().get_tensor_by_name("output:0")
start_time = time()
preds = sess.run(output, feed_dict={inputdata: padded_images})
end_time = time()
print("Recognition time: {}".format(end_time - start_time))
def load_images(image_path: str, files_limit: int):
onlyfiles = [join(image_path, f) for f in listdir(image_path) if isfile(join(image_path, f))][:files_limit]
return np.array([load_and_resize_image(p) for p in onlyfiles]), onlyfiles
def load_images(onlyfiles, files_limit: int):
return np.array([load_and_resize_image(p) for p in onlyfiles]), onlyfiles