-
Notifications
You must be signed in to change notification settings - Fork 1
/
main.py
122 lines (100 loc) · 4.73 KB
/
main.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
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
# 常用资源库
import pandas as pd
import numpy as np
EPS = 1e-9#
import os,glob,numbers
# 图像处理
import math,cv2,random
from PIL import Image, ImageFile, ImageOps, ImageFilter
ImageFile.LOAD_TRUNCATED_IMAGES = True
# 图像显示
from matplotlib import pyplot as plt
plt.rcParams['image.cmap'] = 'gray'
import torch
import torch.nn as nn
import torch.nn.functional as F
from data import *
from nets import *
from build import *
from utils import *
from grad import *
from loop import *
#start#
import argparse
def str2bool(v):
if v.lower() in ('yes', 'true', 't', 'y', '1'):
return True
elif v.lower() in ('no', 'false', 'f', 'n', '0'):
return False
else:
raise argparse.ArgumentTypeError('Unsupported value encountered.')
parser = argparse.ArgumentParser(description="Train network")
# 实验参数 C:对比学习、P：先验知识
parser.add_argument('--inc', type=str, default='', help='instruction')#skeleton & dilation
parser.add_argument('--gpu', type=int, default=0, help='cuda number')
parser.add_argument('--los', type=str, default='fr', help='loss function')
parser.add_argument('--net', type=str, default='lunet', help='network')
parser.add_argument('--seg', type=str, default='lunet', help='network')
# parser.add_argument('--patch', type=str2bool, default=True, help='Patch based!')
parser.add_argument('--csm', type=str2bool, default=False, help='Color Space Mixture!')
parser.add_argument('--coff_ds', type=float, default=0.5, help='Cofficient of Deep Supervision!')
parser.add_argument('--coff_ce', type=float, default=0.1, help='Cofficient of DMF entropy!')
parser.add_argument('--coff_rot', type=float, default=0.005, help='Cofficient of regualar rotation!')
parser.add_argument('--bug', type=str2bool, default=False, help='debug!')
parser.add_argument('--board', type=str2bool, default=True, help='debug!')
# 数据参数
parser.add_argument('--db', type=str, default='stare', help='instruction')
parser.add_argument('--loo', type=int, default=20, help='Leave One Out')
parser.add_argument('--bs', type=int, default=32, help='batch size')
parser.add_argument('--ds', type=int, default=128, help='data size')
parser.add_argument('--pl', type=str2bool, default=False, help='Parallel!')
parser.add_argument('--root', type=str, default='', help='root folder')
# 正则化参数
parser.add_argument('--ct', type=str2bool, default=True, help='Constraint for Network!')
parser.add_argument('--coff_ct', type=float, default=.5, help='Cofficient of Constraint!')
parser.add_argument('--loss_ct', type=str, default='di', help='Loss of Contrastive learning!')
# 其他参数
parser.add_argument('--arch', type=str, default='siam', help='architechture')
parser.add_argument('--roma', type=str2bool, default=False, help='Random Mapping!')
parser.add_argument('--coff_cl', type=float, default=.1, help='Cofficient of Contrastive learning!')
parser.add_argument('--temp_cl', type=float, default=.1, help='Temperature of Contrastive learning!')
parser.add_argument('--loss_cl', type=str, default='sim3', help='Loss of Contrastive learning!')#, choices=['', 'au', 'nce', 'sim', 'nce2', 'sim2']
parser.add_argument('--sss', type=str, default='half', choices=['', 'hard', 'half'], help='Sample Selection Strategy!')
parser.add_argument('--top', type=int, default=4, help='sampler top')
parser.add_argument('--low', type=int, default=2, help='sampler low')
parser.add_argument('--dis', type=int, default=4, help='sampler dis')
parser.add_argument('--num', type=int, default=512, help='sampler number')
args = parser.parse_args()
os.environ["CUDA_DEVICE_ORDER"] = "PCI_BUS_ID"
# os.environ["CUDA_VISIBLE_DEVICES"] = "0,1,2,3"#
os.environ["CUDA_VISIBLE_DEVICES"] = ','.join([str(i) for i in range(args.gpu, 4)])
# 训练程序########################################################
if __name__ == '__main__':
dataset = EyeSetGenerator(dbname=args.db, datasize=args.ds, loo=args.loo)
# dataset = EyeSetGenerator(dbname=args.db, isBasedPatch=args.patch)
dataset.use_csm = args.csm
net = build_model(args.net, args.seg, args.loss_cl, args.arch)
if args.db=='stare' and args.loo<20:
net.__name__ += 'LOO'+str(args.loo)
if args.db in ['chase', 'hrf']:
# dataset.SIZE_IMAGE = args.ds*2
net.encoder.flag_down = True
print('$'*64, 'FLAG_DOWN')
print('$'*64, 'FLAG_DOWN')
keras = KerasTorch(model=net, args=args)
keras.args = args
keras.isParallel = args.pl
if args.ct or 'dmf' in net.__name__:
args.ct = True
net.__name__ += args.loss_ct + str(args.coff_ct)
keras.loss_ct = get_loss(args.loss_ct)
else:
args.ct = False
net.__name__ += args.inc + 'ds'+str(args.coff_ds) + args.sss
print('Network Name:', net.__name__)
keras.compile(dataset, loss=args.los, lr=0.01)
keras.gradUtil.coff_ds = args.coff_ds
if args.root=='':
keras.val()
keras.fit(epochs=169)
#end#