def score(root: Tk, user_name: str, image_name: str, image_number: str):
image_dir_path = root_image_dir_path / image_name / image_number / 'random_enhance_10'
if not image_dir_path.exists():
raise FileNotFoundError
elif not image_dir_path.is_dir():
raise NotADirectoryError
game = TournamentGame(list(map(str, image_dir_path.iterdir())),
ImageGenerator())
scored_image_dir = get_save_dir_path(root_scored_image_dir_path, user_name,
f'{image_name}/{image_number}')
scored_param_dir = get_save_dir_path(root_scored_param_dir_path, user_name,
f'{image_name}/{image_number}')
scored_param_path = get_save_file_path(scored_param_dir,
'scored_param.json')
data_writer_list = [
DW_scored_image.ScoredImageWriter(str(scored_image_dir)),
DW_scored_param.ScoredParamWriter(str(scored_param_path))
]
sub_win = Toplevel(root)
canvas = CompareCanvasGroupFrame(sub_win,
game,
data_writer_list=data_writer_list)
canvas.pack()
canvas.disp_enhanced_image()
canvas.focus_set()
sub_win.grab_set()
sub_win.wait_window()
def make_train_data(root: tk.Tk, user_name: str,
image_name: str) -> (bool, str):
image_dir_path = config.path.root_image_dir_path / \
image_name/'1'
if not image_dir_path.exists():
raise MakeTrainDataException(f'{str(image_dir_path)} is not found')
elif not image_dir_path.is_dir():
raise MakeTrainDataException(f'{str(image_dir_path)} is not directory')
image_path = list(
itertools.chain(image_dir_path.glob('*.jpg'),
image_dir_path.glob('*.png')))[0]
try:
scored_param_dir_path = get_save_dir_path(
config.path.root_scored_param_dir_path, user_name, image_name)
scored_param_path = get_save_file_path(
scored_param_dir_path,
'scored_param' + DataWriter.ScoredParamWriter.SUFFIX)
except MiscException as e:
raise MakeTrainDataException(e)
sub_win = tk.Toplevel(root)
try:
compare_num = 100
is_complete = compare(sub_win, image_path, scored_param_path,
compare_num)
return is_complete, scored_param_path
except TrainDataMakerException as e:
sub_win.destroy()
raise MakeTrainDataException(e)
def get_summary_dir_path_func():
return get_save_dir_path(
root_summary_dir_path, user_name, image_name)
scored_param_list = None
with open(args.scored_param_path, 'r') as fp:
scored_param_list = json.load(fp)
image_list = [{
'image': Image.open(scored_param['param'])
} for scored_param in scored_param_list]
evaluate_list = predict_model.predict(image_list)
for i in range(len(scored_param_list)):
scored_param_list[i]['evaluate'] = evaluate_list[i][0]
rightfulness_dir_path = get_save_dir_path(
root_rightfulness_dir_path, args.user_name,
f'{args.image_name}/{args.image_number}')
scored_param_list.sort(key=lambda x: x['score'], reverse=True)
for index, scored_param in enumerate(tqdm(scored_param_list)):
image = Image.open(scored_param['param'])
new_image = Image.new(image.mode, (image.width, image.height + 100),
(255, 255, 255))
new_image.paste(image, (0, 0))
draw = ImageDraw.Draw(new_image)
text_dict = {
key: f'{key:<10s}:{scored_param[key]:>5,.2f}'
def save_dir_path_func():
image_number = self.image_file_path.parent.name
return get_save_dir_path(root_optimize_dir_path, self.user_name,
f'{self.image_name}/{image_number}')
color='r', label="Average Fitness")
ax1.plot(log_dict['gen'], log_dict['min'],
color='g', label="Minimum Fitness")
ax1.plot(log_dict['gen'], log_dict['max'],
color='b', label="Maximum Fitness")
ax1.legend()
save_file_path = get_save_file_path(log_dir_path, 'graph.png')
plt.savefig(str(save_file_path))
if __name__ == "__main__":
args = _get_args()
optimize_dir_path = get_save_dir_path(
root_optimize_dir_path, args.user_name, f'{args.image_name}/{args.image_number}')
image_dir = root_image_dir_path/args.image_name/args.image_number
image_path = list(image_dir.glob('*.*'))[0]
enhancer = ResizableEnhancer(str(image_path), IMAGE_SIZE)
model = RankNet(IMAGE_SHAPE)
model.load(args.weights_file_path)
image_generator = EnhanceGenerator(enhancer)
best_param_list, _ = Optimizer(model, image_generator, EnhanceDecorder()).optimize(20)
for index, best_param in enumerate(best_param_list):
save_path = str(Path(optimize_dir_path)/f'best_{index}.png')
enhancer.enhance(best_param).save(save_path)