e2_main.py

import argparse, os
os.environ['TF_CPP_MIN_LOG_LEVEL'] = '1' 
from e2_model import model

parser	= argparse.ArgumentParser(description='')
parser.add_argument('--phase',		dest='phase',					default='test',		help='train or test')
parser.add_argument('--patches',	dest='num_patch',	type=int,	default=4,			help='number of patches extracted from an image')
parser.add_argument('--patchsize',	dest='patch_sz',	type=int,	default=64,			help='patch size, height/width of frames in each burst')
parser.add_argument('--burstsize',	dest='burst_sz',	type=int,	default=8,			help='burst size, number of frames in each burst')
parser.add_argument('--batchsize',	dest='batch_sz',	type=int,	default=4,			help='batch size, number of bursts in each batch')

parser.add_argument('--epochs',		dest='epochs',		type=int,	default=200,		help='number of epoch')
parser.add_argument('--lr',			dest='lr',			type=float,	default=0.001,		help='learning rate')

parser.add_argument('--jit',		dest='jit',			type=int,	default=2,			help='jit')
parser.add_argument('--J',			dest='J',			type=int,	default=2,			help='downsample ratio J')
parser.add_argument('--alpha',		dest='alpha',		type=float,	default=4.0,		help='alpha')

parser.add_argument('--ckpt_dir',	dest='ckpt_dir',				default='./e2_ckpt',							help='the directory that models are saved')
parser.add_argument('--train_dir',	dest='train_dir',				default='./Datasets/VOC2008/train',				help='the directory that training images are saved')
parser.add_argument('--valid_dir',	dest='valid_dir',				default='./Datasets/VOC2008/val',				help='the directory that validation images are saved')
parser.add_argument('--test_dir',	dest='test_dir',				default='./Datasets/VOC2008/test',				help='the directory that test images are saved')
parser.add_argument('--output_dir',	dest='output_dir',				default='./Results/VOC2008/e2_out',				help='the directory to save processed test images')
parser.add_argument('--ckpt_file',	dest='ckpt_file',				default='model.h5',								help='the file that models are saved')

parser.add_argument('--shift_ckpt',	dest='shift_ckpt',				default='./shift_ckpt/model.h5',				help='the file of shift model')
parser.add_argument('--noisy_ckpt',	dest='noisy_ckpt',				default='./noisy_ckpt/model.h5',				help='the file of noisy model')

parser.add_argument('--seed',		dest='seed',		type=int,	default=42,			help='random seed number')
args	= parser.parse_args()
#70


def main():
	if not os.path.exists(args.ckpt_dir):
		os.makedirs(args.ckpt_dir)

	model_ = model(args)
	if args.phase == 'train':
		model_.train()
	else:
		if not os.path.exists(args.output_dir):
			os.makedirs(args.output_dir)
		model_.test()


if __name__ == "__main__":
	main()