-
Notifications
You must be signed in to change notification settings - Fork 5
/
params.py
31 lines (23 loc) · 1.3 KB
/
params.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
import numpy as np
import os
import torch
def get_params():
params = {}
params['experiment_log'] = "MTL_ser_r101u_0" # experiment ID.
params['main_dir'] = r"C:\Users\burak\Desktop"
params['log_path'] = os.path.join(params['main_dir'], 'mtl', 'logs') # path to save logs.
params['main_data_dir'] = r"D:\Veri-Setleri\SN6\SN6_buildings_AOI_11_Rotterdam_train\train\AOI_11_Rotterdam"
params['masks_dir'] = os.path.join(params['main_data_dir'], 'MASKS_binary') # dataset path.
params['psp_dir'] = os.path.join(params['main_data_dir'], 'PS-RGB') # dataset path.
params['crop_size'] = (480,480)
params['encoder'] = 'resnet101' # backbone architecture, encoder.
params['encoder_weights'] = 'imagenet' # pre-trained weights.
params['classes'] = np.arange(0,2,1) # used for encoding-decoding the mask.['building','not building']
params['activation'] = 'sigmoid' # activation function.
params['device'] = torch.device("cuda" if torch.cuda.is_available() else "cpu") # GPU or CPU.
params['batch_size'] = 4 # batch size.
params['lr'] = 0.0001 # learning rate.
params['n_epoch'] = 50 # number of epochs.
params['n_workers'] = 0 #number of workers for data loader, multi-process scheme.
params['seed'] = 12
return params