options.py

² import argparse


parser = argparse.ArgumentParser()
parser.add_argument('--epoch',       type=int,   default=100,   help='epoch number')
parser.add_argument('--lr',          type=float, default=1e-4,  help='learning rate')
parser.add_argument('--batchsize',   type=int,   default=8,    help='training batch size')
parser.add_argument('--trainsize',   type=int,   default=512,   help='training dataset size')
parser.add_argument('--clip',        type=float, default=0.5,   help='gradient clipping margin')
parser.add_argument('--lw',          type=float, default=0.001, help='weight')
parser.add_argument('--decay_rate',  type=float, default=0.1,   help='decay rate of learning rate')
parser.add_argument('--decay_epoch', type=int,   default=60,    help='every n epochs decay learning rate')
#parser.add_argument('--load',        type=str,   default='./SPNet_epoch_best.pth',  help='train from checkpoints') # if continue learning from previous stage
parser.add_argument('--load',        type=str,   default='',  help='train from checkpoints')

parser.add_argument('--gpu_id',      type=str,   default='1',   help='train use gpu')

parser.add_argument('--rgb_label_root',      type=str, default='./COD-TrainDataset/Imgs/',           help='the training rgb images root')
parser.add_argument('--depth_label_root',    type=str, default='./COD-TrainDataset/depth/',         help='the training depth images root')
parser.add_argument('--gt_label_root',       type=str, default='./COD-TrainDataset/GT/',            help='the training gt images root')

# or use other datasets for validation
parser.add_argument('--val_rgb_root',        type=str, default='./COD-TestDataset/CAMO/Imgs/',      help='the test rgb images root')
parser.add_argument('--val_depth_root',      type=str, default='./COD-TestDataset/CAMO/depth/',    help='the test depth images root')
parser.add_argument('--val_gt_root',         type=str, default='./COD-TestDataset/CAMO/GT/',       help='the test gt images root')

parser.add_argument('--save_path',           type=str, default='./Checkpoint/SPNet/',    help='the path to save models and logs')

opt = parser.parse_args()