HumanSkinSegmentation/utils.py at master · gbirman/HumanSkinSegmentation · GitHub

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
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143

from pathlib import Path
import os
from distutils.dir_util import copy_tree

def getArgs(train=True):
    import argparse

    # determine if training on local or remote
    parser = argparse.ArgumentParser()
    parser.add_argument('-lo', '--local', action="store_true", dest="local")
    args = parser.parse_known_args()[0]

    # default settings
    fname = "BGR"
    data_root="./Data" # where is the data stored
    just_subtract = False
    lr = 1e-3
    wd = 2e-4
    frame_num = 5
    load_init = "BGR_original"
    load_multi = "LIRIS_5_res_wd"
    if train:
        dataset_name = "FSD"
    else:
        dataset_name = "Pratheepan_Face"

    # dataloader settings
    num_workers = 8 if args.local else 32
    pin_memory = False if args.local else True
    if train:
        batch_size = 8 if args.local else 32
        validation_split = 0.3
        shuffle_dataset = True
    else:
        batch_size = 1
        validation_split = 0.3
        shuffle_dataset = False

    # optional command line arguments
    parser.add_argument('-s', '--sigmoid', action="store_true", dest="use_sigmoid")
    parser.add_argument('-g', '--grayscale', action="store_true", dest="grayscale")
    parser.add_argument('-lr', action="store", dest="lr", default=lr, type=float)
    parser.add_argument('-fn', '--file_name', action="store", dest="fname", default=fname, type=str)
    parser.add_argument('-dr', '--data_root', action="store", dest="data_root", default=data_root, type=str)
    parser.add_argument('-dn', '--dataset_name', action="store", dest="dataset_name", default=dataset_name, type=str)
    parser.add_argument('-c', '--create_data_file', action="store_true", dest="createDataFile")
    parser.add_argument('-js', '--just_subtract', action="store_true", dest="just_subtract", default=just_subtract)
    parser.add_argument('-bs', '--batch_size', action="store", dest="batch_size", default=batch_size, type=int)
    parser.add_argument('-ln', '--load_init', action="store", dest="load_init", default=load_init, type=str)
    parser.add_argument('-m', '--multi', action="store_true", dest="multi")
    parser.add_argument('-fr', '--frame_num', action="store", dest="frame_num", default=frame_num, type=int)
    parser.add_argument('-lm', '--load_multi', action="store", dest="load_multi", default=load_multi, type=str)
    parser.add_argument('-o', '--original', action="store_true", dest="original")
    parser.add_argument('-r', '--res', action="store_true", dest="res")
    parser.add_argument('-wd', '--weight_decay', action="store", dest="wd", default=wd, type=float)
    args = parser.parse_args()
    dl_args = {'batch_size': args.batch_size, 'validation_split': validation_split, 'shuffle_dataset': shuffle_dataset,
                'pin_memory': pin_memory, 'num_workers': num_workers}
    print(vars(args), dl_args)

    return args, dl_args

# display figures for evaluation purposes (assumes batch_size = 1)
def save_figure(save_file, inp, tgt, pred, pred_bin, grayscale=False, show=False):
    import matplotlib.pyplot as plt

    # bring to shape: N x C x D x H X W
    if inp.dim() == 4:
        inp = inp.unsqueeze_(2)

    if tgt.dim() == 4:
        tgt = tgt.unsqueeze_(2)

    if pred.dim() == 4:
        pred = pred.unsqueeze_(2)

    if pred_bin.dim() == 4:
        pred_bin = pred_bin.unsqueeze_(2)

    # bring to shape: C x D x H X W
    inp = inp.squeeze(0)
    tgt = tgt.squeeze(0)
    pred = pred.squeeze(0)
    pred_bin = pred_bin.squeeze(0)

    # show total
    _, axs = plt.subplots(inp.shape[1], 4, frameon=False) # D x 4
    if inp.shape[1] == 1:
        axs = [axs]
    axs[0][0].set_title('Input')
    axs[0][1].set_title('Target')
    axs[0][2].set_title('Predicted')
    axs[0][3].set_title('Binarized')
    for i in range(inp.shape[1]):
        inp_img = inp[:,i,:,:].permute(1,2,0).flip(-1).squeeze(2).cpu().numpy()
        axs[i][0].imshow(inp_img)
        if i == inp.shape[1]//2:
            tgt_img = tgt[:,0,:,:].permute(1,2,0).squeeze(2).cpu().numpy()
            pred_img = pred[:,0,:,:].permute(1,2,0).squeeze(2).cpu().numpy()
            pred_bin_img = pred_bin[:,0,:,:].permute(1,2,0).squeeze(2).cpu().numpy()
            axs[i][1].imshow(tgt_img)
            axs[i][2].imshow(pred_img)
            axs[i][3].imshow(pred_bin_img)
        axs[i][0].axis('off')
        axs[i][1].axis('off')
        axs[i][2].axis('off')
        axs[i][3].axis('off')
    plt.savefig(f"{save_file}.png")
    if show:
        plt.show()
    plt.close()

    # show individual
    for i in range(inp.shape[1]):
        inp_img = inp[:,i,:,:].permute(1,2,0).flip(-1).squeeze(2).cpu().numpy()
        plt.imshow(inp_img)
        plt.axis('off')
        plt.savefig(f"{save_file}inp_{i}.png")
        plt.close()
        if i == inp.shape[1]//2:
            tgt_img = tgt[:,0,:,:].permute(1,2,0).squeeze(2).cpu().numpy()
            pred_img = pred[:,0,:,:].permute(1,2,0).squeeze(2).cpu().numpy()
            pred_bin_img = pred_bin[:,0,:,:].permute(1,2,0).squeeze(2).cpu().numpy()
            plt.imshow(tgt_img)
            plt.axis('off')
            plt.savefig(f"{save_file}tgt.png")
            plt.close()
            plt.imshow(pred_img)
            plt.axis('off')
            plt.savefig(f"{save_file}pred.png")
            plt.close()
            plt.imshow(pred_bin_img)
            plt.axis('off')
            plt.savefig(f"{save_file}bin.png")
            plt.close()

# log and print info simultaneously
def log(msg, logger, print_msg=True, log_msg=True):
    if print_msg:
        print(msg)
    if log_msg:
        logger.error(msg)