def iterative_prediction(image_path, image_name, result_image_path,
result_path, iterative_time):
#copy original image to result_image_path
if not os.path.exists(result_image_path):
os.mkdir(result_image_path)
if not os.path.exists(result_path):
os.mkdir(result_path)
image_name, image_ext = os.path.splitext(image_name)
shutil.copyfile(os.path.join(image_path, image_name+image_ext),
os.path.join(result_image_path, image_name+'+0'+image_ext))
results = []
#establish network
east = East()
east_detect = east.east_network()
east_detect.load_weights(cfg.saved_model_weights_file_path)
#start iteration
for i in range(1, iterative_time + 1, 1):
#load i-th result as input to predict
current_result = predict(east_detect,
os.path.join(result_image_path,image_name + '+' +
str(i-1)+image_ext) ,
text_pixel_threshold= cfg.text_pixel_threshold - i *
0.05,
text_side_threshold=
cfg.text_side_vertex_pixel_threshold - i * 0.05,
text_trunc_threshold= cfg.text_trunc_threshold +
i * 0.03,
action_pixel_threshold=
cfg.action_pixel_threshold,
arrow_trunc_threshold=
cfg.arrow_trunc_threshold,
nock_trunc_threshold=
cfg.nock_trunc_threshold,
quiet=False)
results.extend(current_result)
#according to current results, erase last iteration image
erase_image(current_result,os.path.join(result_image_path,image_name + '+' +
str(i-1)+image_ext))
del current_result
with open(os.path.join(result_path, image_name+'.txt'),'w') as result_fp:
result_fp.writelines(results)
del results
def main(img_path):
east = East()
east_detect = east.east_network()
east_detect.load_weights(cfg.saved_model_weights_file_path)
for i in range(int(cfg.max_self_iteration)):
predict(east_detect, img_path, i, quiet=True)
#load predicted images and results
with open(img_path[:-4] + '.txt') as result_fp:
results = result_fp.readlines()
if i != 0:
image_path = img_path[:-4] + '_' + str(i - 1) + '.png'
else:
image_path = img_path
input_img = cv2.imread(image_path)
ori_img = cv2.imread(img_path)
#with Image.open(image_path) as im:
#im_array = im.img_to_array()
# draw = ImageDraw.Draw(im.convert('RGB'))
for line in results:
geo = line.split(',')
# erase detective text
cv2.rectangle(input_img, (int(
(geo[0])) - cfg.erase_offset_pixels, int(
(geo[1])) - cfg.erase_offset_pixels), (int(
(geo[4])) + cfg.erase_offset_pixels, int(
(geo[5])) + cfg.erase_offset_pixels),
(255, 255, 255),
thickness=-1)
#marked predict results
cv2.rectangle(ori_img, (int((geo[0])), int((geo[1]))), (int(
(geo[4])), int((geo[5]))), (255, 0, 0),
thickness=1)
cv2.imwrite(img_path[:-4] + '_' + str(i) + '.png', input_img)
cv2.imwrite(img_path[:-4] + '_predict_' + str(i) + '.png', ori_img)
def train_model():
east = East()
east_network = east.east_network()
east_network.summary()
east_network.compile(
loss=quad_loss,
optimizer=Nadam(
lr=cfg.lr,
# clipvalue=cfg.clipvalue,
schedule_decay=cfg.decay))
# load pre-trained model
if cfg.load_weights and os.path.exists(cfg.saved_model_weights_file_path):
east_network.load_weights(cfg.saved_model_weights_file_path,
by_name=True,
skip_mismatch=True)
print('start training task:' + cfg.train_task_id + '....................')
# train on current scale data
east_network.fit_generator(
generator=gen(),
steps_per_epoch=cfg.steps_per_epoch,
epochs=cfg.epoch_num,
validation_data=gen(is_val=True),
validation_steps=cfg.validation_steps,
verbose=2,
initial_epoch=cfg.initial_epoch,
callbacks=[
EarlyStopping(patience=cfg.patience, verbose=2),
ModelCheckpoint(filepath=cfg.saved_model_weights_file_path,
save_best_only=True,
save_weights_only=True,
verbose=1)
])
del east_network, east
def parse_args():
parser = argparse.ArgumentParser()
parser.add_argument('--path', '-p',
default='../data/003.jpg',
help='image path')
parser.add_argument('--threshold', '-t',
default=cfg.pixel_threshold,
help='pixel activation threshold')
return parser.parse_args()
if __name__ == '__main__':
args = parse_args()
img_path = args.path
threshold = float(args.threshold)
print(img_path, threshold)
img = image.load_img(img_path)
im_name = img_path.split('/')[-1][:-4]
east = East()
east_detect = east.east_network()
east_detect.load_weights(cfg.saved_model_weights_file_path)
predict(east_detect, img_path, threshold)
# text_recs_all, text_recs_len, img_all = predict_quad(east_detect, img, img_name=im_name)
# print(text_recs_all)
# print("-------------------------")
# print(text_recs_len)
# print("-------------------------")
# print(img_all)
# img = image.array_to_img(img_all[0])
# img.show()
import os
from keras.callbacks import EarlyStopping, ModelCheckpoint
from keras.optimizers import Adam
import cfg
from network import East
from losses import quad_loss
from data_generator import gen
east = East()
east_network = east.east_network()
east_network.summary()
east_network.compile(loss=quad_loss, optimizer=Adam(lr=cfg.lr,
# clipvalue=cfg.clipvalue,
decay=cfg.decay))
if cfg.load_weights and os.path.exists(cfg.saved_model_weights_file_path):
east_network.load_weights(cfg.saved_model_weights_file_path)
east_network.fit_generator(generator=gen(),
steps_per_epoch=cfg.steps_per_epoch,
epochs=cfg.epoch_num,
validation_data=gen(is_val=True),
validation_steps=cfg.validation_steps,
verbose=1,
initial_epoch=cfg.initial_epoch,
callbacks=[
EarlyStopping(patience=cfg.patience, verbose=1),
ModelCheckpoint(filepath=cfg.model_weights_path,
save_best_only=True,
save_weights_only=True,
verbose=1)])
def start_prediction(cropped_path):
east = East()
east_detect = east.east_network()
east_detect.load_weights("./east_model_weights_3T736.h5")
predict(east_detect, cropped_path, cfg.pixel_threshold)
