SPECIAL/test.py at main · LiPang/SPECIAL · 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
import os
import argparse
import random
import warnings

import h5py
import torch
import numpy as np
from scipy.io import *
from utils.utility import rsshow
from hyperseg.HyperSeg import hyperseg_classification
import spectral

# Set environment variables and warnings
warnings.filterwarnings('ignore')

def seed_everywhere(seed=42):
    """Set the random seed for reproducibility across all devices."""
    torch.manual_seed(seed)
    if torch.cuda.is_available():
        torch.cuda.manual_seed(seed)
        torch.cuda.manual_seed_all(seed)
    torch.backends.cudnn.deterministic = True
    torch.backends.cudnn.benchmark = False
    np.random.seed(seed)
    random.seed(seed)

def load_paviac():
    label_list = [
        'Background',
        'Water',
        'Trees',
        'Meadows',
        'Self-Blocking Bricks',
        'Bare Soil',
        'Asphalt',
        'Bitumen',
        'Tile',
        'Shadows'
    ]
    data_name = 'PaviaC'
    img_ori = loadmat("data/PaviaCentre/Pavia.mat")['pavia']
    label_gt = loadmat("data/PaviaCentre/Pavia_gt.mat")['pavia_gt']
    img_rgb = loadmat("data/PaviaCentre/data_mat_rgb.mat")['data']

    positions = [(0, 1096, 0, 715)]

    label_copy = label_gt.copy()
    label_list = [label_list[int(t)] for t in np.unique(label_gt)]
    for t, tt in enumerate(np.unique(label_gt)):
        if label_gt.min() > 0:
            label_copy[label_gt == tt] = t + 1  # no background
        else:
            label_copy[label_gt == tt] = t
    if label_gt.min() == 0:
        label_list.pop(0)
    label_gt = label_copy
    return data_name, img_ori, label_gt, label_list, positions, img_rgb


def load_aerorit():
    label_list = ["Background", "Buildings", "River or Lake", "Cars",
                  "Grass or Trees", "Soil or Road"]
    data_name = 'AeroRIT'
    data = spectral.open_image(
        "data/AeroRIT/Reflectance/2017-08-01_14-28-56_000011__061158-066258_radiance_fwd_proj_EmpLine.hdr")
    img_ori_ = np.array(data[:, :, :])
    img_ori = img_ori_[..., 10:-10]

    label_gt = loadmat("data/AeroRIT/label_mat.mat")['data']
    mapping = np.array([0, 1, 4, 5, 2, 3])
    label_gt = mapping[label_gt]
    img_rgb = loadmat("data/AeroRIT/data_mat_rgb.mat")['data']

    img_ori = img_ori[500:500+1024, 100:100+1024*3]
    label_gt = label_gt[500:500+1024, 100:100+1024*3]
    img_rgb = img_rgb[500:500+1024, 100:100+1024*3]
    positions = []
    gap = 512
    for x in range(0, 1024, gap):
        for y in range(0, 1024 * 3, gap):
            positions.append((x, x + gap, y, y + gap))


    label_copy = label_gt.copy()
    label_list = [label_list[int(t)] for t in np.unique(label_gt)]
    for t, tt in enumerate(np.unique(label_gt)):
        if label_gt.min() > 0:
            label_copy[label_gt == tt] = t + 1  # no background
        else:
            label_copy[label_gt == tt] = t
    if label_gt.min() == 0:
        label_list.pop(0)
    label_gt = label_copy
    return data_name, img_ori, label_gt, label_list, positions, img_rgb


def build_settings():
    parser = argparse.ArgumentParser("")
    parser.add_argument('--learning_rate', type=float, default=1e-3, help='learning rate')
    parser.add_argument('--eta_min', type=float, default=1e-4, help='learning rate')
    parser.add_argument('--weight_decay', type=float, default=0.0, help='weight_decay')
    parser.add_argument('--iterations', type=int, default=20, help='learning iterations')
    parser.add_argument('--warmup_epoches', type=int, default=10)
    parser.add_argument('--epoches', type=int, default=20, help='epoch number')
    parser.add_argument('--samples', type=int, default=500)
    parser.add_argument('--data_name', type=str, default='AeroRIT')
    parser.add_argument('--lambda1', type=float, default=0.2)
    parser.add_argument('--lambda2', type=float, default=1)
    args = parser.parse_args()
    return args

if __name__ == '__main__':
    args = build_settings()

    seed_everywhere(42)

    if args.data_name == 'PaviaC':
        data_name, image_hsi, label_gt, label_list, positions, image = load_paviac()
        args.data_name = data_name
    elif args.data_name == 'AeroRIT':
        data_name, image_hsi, label_gt, label_list, positions, image = load_aerorit()
        args.data_name = data_name

    image_hsi = rsshow(image_hsi, 0.0001)
    image = rsshow(image, 0.001)

    hyperseg_classification(image, image_hsi, label_list, label_gt, positions, args)
    print('ok')