-
Notifications
You must be signed in to change notification settings - Fork 2
/
config.py
36 lines (36 loc) · 1.71 KB
/
config.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
model ='ours'#'PNN'#'PanNet'#'TFNet'##'SRPPNN'# # deep learning pan-sharpening method such as PNN MSDCNN PanNet TFNet SRPPNN and DIPNet
model_sub = 'F64B2'# for our DIPNet's ablation study, such as SR SR_PLB SR_PHB SR_PLB_PHB default set as ''
model_loss = ''# for our DIPNet's ablation study, such as SSIM or L1 or '', default set as ''
isTrain = True #whether or not to train
gpu_ids = '0'
continue_train = False# whether or not to continue
which_epoch = -1 # if continue Train ture, set this value
print_net_in_detail = False
dataDir = r'D:\实验\影像融合\IKONOS\train_dataset'# this is a train dataset dir
testdataDir = r'D:\实验\影像融合\IKONOS\test_dataset' # this is a test dataset dir
save_result = True # whether or not to save result
saveDir = r'D:\实验\影像融合\Deep-Learning-PanSharpening\results\IKONOS\result'# save directory
checkpoints_dir = r'D:\实验\影像融合\Deep-Learning-PanSharpening\checkpoints\IKONOS\checkpoints_dir'# the dir of pertrained models or the dir to checkpoint the model parameters during training
nThreads = 0 # use serval threads to load data in pytorch
batchSize = 16
img_size = 32 # simulate ms size
scale = 4 # scale factor which resizing to pan size
seed = 19970716 # random seed
print_freq = 5 # print frequency of log
save_epoch_freq = 100
pan_channel = 1 # pan-chromatic band
mul_channel = 4 # multi-spectral band which is based on different satellite
gan_mode = 'lsgan' #'lsgan' or 'wgangp' or 'vanilla' this is orginal
data_range = 2047 # radis resolution
lr_policy = 'step'
optim_type = 'adam'#'adam'# c
lr = 1e-4
beta = 0.9
momentum = 0.9
weight_decay = 1e-8
lr_decay_iters = [1000]
lr_decay_factor = 0.5
epochs = 1000
isUnlabel = False
isEval = True#False
useFakePAN = False