-
Notifications
You must be signed in to change notification settings - Fork 3
/
config.py
31 lines (29 loc) · 1.43 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
##########################################
# @subject : Person segmentation #
# @author : perryxin #
# @date : 2018.12.27 #
##########################################
import numpy as np
class Config():
def __init__(self):
self.PATH = "/data1/datasets/supervisely/"
self.PATH_VOC = "/data1/datasets/VOCdevkit/VOC2012"
self.PATH_COCO = "/data1/datasets/COCO2017"
self.PATH_VIP = "/data1/datasets/VIP_Fine"
self.PATH_ATR = "/data1/datasets/LIP/ATR/humanparsing"
self.PATH_CHIP = "/data1/datasets/LIP/CIHP/instance-level_human_parsing"
self.PATH_LIP = "/data1/datasets/LIP/LIP/trainval"
self.PATH_MHP = "/data1/datasets/LV-MHP-v2/LV-MHP-v2"
self.PATH_TRIMODAL = "/data1/datasets/trimodal"
self.PATH_SIT = "/data1/datasets//SIT"
self.split_train = ["seg__ds1", "seg__ds2", "seg__ds3", "seg__ds4", "seg__ds5", "seg__ds6", "seg__ds7",
"seg__ds8"]
self.split_test = ["seg__ds9", "seg__ds10", "seg__ds11", "seg__ds12", "seg__ds13"]
self.mean = np.array([[[0.485]], [[0.456]], [[0.406]]]).transpose(1, 2, 0)
self.std = np.array([[[0.229]], [[0.224]], [[0.225]]]).transpose(1, 2, 0)
self.BATCH_SIZE_TRAIN = 16 # 4
self.BATCH_SIZE_VAL = 1
self.BATCH_SIZE_TEST = 1
self.EPOCH = 300
self.WEIGHT_DECAY = 10 ** -7
self.LR = 10 ** -3 # 0.001