config.py

"""
    All of the parameters are defined here.
"""


import time
import argparse

parser = argparse.ArgumentParser(description='RGBD fusion')

# Dataset
parser.add_argument('--data_dir',
                    type=str,
                    default='../data/train',
                    help='path to dataset')
parser.add_argument('--data_list',
                    type=str,
                    default='data.list',
                    help='data list')
parser.add_argument('--patch_height',
                    type=int,
                    default=256,
                    help='height of a patch to crop')
parser.add_argument('--patch_width',
                    type=int,
                    default=256,
                    help='width of a patch to crop')


# Hardware
parser.add_argument('--seed',
                    type=int,
                    default=7240,
                    help='random seed point')
parser.add_argument('--num_gpus',
                    type=int,
                    default=1,
                    help='number of visible GPUs')
parser.add_argument('--port',
                    type=str,
                    default='3009',
                    help='multiprocessing port')
parser.add_argument('--num_threads',
                    type=int,
                    default=1,
                    help='number of threads')


# Training
parser.add_argument('--loss',
                    type=str,
                    default='1.0*L1+1.0*L2',
                    help='loss function configuration')
parser.add_argument('--pretrain',
                    type=str,
                    default=None,
                    help='path of pretrained model')
parser.add_argument('--epochs',
                    type=int,
                    default=100,
                    help='number of epochs to train')
parser.add_argument('--batch_size',
                    type=int,
                    default=20,
                    help='input batch size for training')
parser.add_argument('--max_depth',
                    type=float,
                    default=10.0,
                    help='maximum depth')
parser.add_argument('--depth_norm',
                    action='store_true',
                    default=True,
                    help='Normalize spot depth')
parser.add_argument('--augment',
                    type=bool,
                    default=True,
                    help='data augmentation')
parser.add_argument('--num_spot',
                    type=int,
                    default=1000,
                    help='number of sparse depth spots')
parser.add_argument('--grid_spot',
                    action='store_true',
                    default=True,
                    help='random spots by grid')
parser.add_argument('--cut_mask',
                    action='store_true',
                    default=False,
                    help='randomly cut parts of the input spots')
parser.add_argument('--noise',
                    type=float,
                    default=0.0,
                    help='stddev of guassian noise augmentation')
parser.add_argument('--rgb_noise',
                    type=float,
                    default=0.0,
                    help='stddev of rgb guassian noise augmentation')

# Testing
parser.add_argument('--test_only',
                    action='store_true',
                    default=False,
                    help='test only flag')
parser.add_argument('--test_name',
                    type=str,
                    default='test',
                    help='file name for saving testing results')


# Summary
parser.add_argument('--num_summary',
                    type=int,
                    default=5,
                    help='maximum number of summary images to save')

# Optimizer
parser.add_argument('--lr',
                    type=float,
                    default=0.001,
                    help='learning rate')
parser.add_argument('--decay',
                    type=str,
                    default='10,20,30,40,50',
                    help='learning rate decay schedule')
parser.add_argument('--freeze_bn',
                    action='store_true',
                    default=False,
                    help='freeze bn')
parser.add_argument('--gamma',
                    type=str,
                    default='1.0,0.5,0.25,0.125,0.0625',
                    help='learning rate multiplicative factors')
parser.add_argument('--optimizer',
                    default='ADAM',
                    choices=('SGD', 'ADAM', 'RMSprop','ADAMW'),
                    help='optimizer to use (SGD | ADAM | RMSprop)')
parser.add_argument('--momentum',
                    type=float,
                    default=0.9,
                    help='SGD momentum')
parser.add_argument('--betas',
                    type=tuple,
                    default=(0.9, 0.999),
                    help='ADAM beta')
parser.add_argument('--epsilon',
                    type=float,
                    default=1e-8,
                    help='ADAM epsilon for numerical stability')
parser.add_argument('--weight_decay',
                    type=float,
                    default=0.0,
                    help='weight decay')
parser.add_argument('--warm_up',
                    action='store_true',
                    default=True,
                    help='do lr warm up during the 1st epoch')

# Logs
parser.add_argument('--save',
                    type=str,
                    default='baseline',
                    help='file name to save')
parser.add_argument('--save_full',
                    action='store_true',
                    default=False,
                    help='save optimizer, scheduler and amp in '
                         'checkpoints (large memory)')
parser.add_argument('--save_image',
                    action='store_true',
                    default=False,
                    help='save images for test')
parser.add_argument('--save_result_only',
                    action='store_true',
                    default=False,
                    help='save result images only with submission format')


args = parser.parse_args()

current_time = time.strftime('%y%m%d_%H%M%S_')
save_dir = '../experiments/' + args.save
args.save_dir = save_dir