UDRC/test.py at master · PengweeWang/UDRC · 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
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
import argparse
import torch
from tqdm import tqdm
from model.ResNet import ResNet, BasicBlock
from model.Unet import Unet
from model.CascadeModel import CascadeModel
from dataset import RMLDataset, BaseRadioDataset
from torch.utils.data import DataLoader
import numpy as np
import matplotlib.pyplot as plt
from sklearn.metrics import confusion_matrix
import seaborn as sns

device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')

parser = argparse.ArgumentParser(
    description='test program for all model',
)

parser.add_argument('-d', '--dataset', default="./data/data_classifier_time_stft.pkl",
                    help="dataset for classifier", type=str)
parser.add_argument("-b", '--batch_size', default=512, type=int)
parser.add_argument("-t", '--type', choices=['unet-noise', 'unet-signal', 'unet-resnet', 'resnet'], type=str)
parser.add_argument("-m", '--model', type=str)
parser.add_argument("-cm", '--cls_model', default="./checkpoints/resnet.pth", type=str, help='classifier model path')

args = parser.parse_args()
if args.type in ['unet-noise', 'unet-signal'] and not args.cls_model:
    parser.error(f"when testing {args.type} model you need to provide --cls_model param (the path of classifier model)")


def test(testModel, testDataset, SNR_list: list, batch_size = 64) -> tuple:
    testModel.eval()
    testModel.to(device)
    y_list = []
    y_pre_list = []
    accuracy_list = []
    for snr in tqdm(SNR_list):
        test_dataset = testDataset.get_snr_data(snr)
        test_data = DataLoader(test_dataset, batch_size=batch_size, shuffle=True, drop_last=False)
        with torch.no_grad():
            accuracy = 0.0
            for x, y in test_data:
                x, y = x.to(device), y.to(device)
                y_pred = testModel(x)
                y_list.append(y.argmax(dim=1).cpu().numpy())
                y_pre_list.append(y_pred.argmax(dim=1).cpu().numpy())
                accuracy_num = (y.argmax(dim=1) == y_pred.argmax(dim=1)).sum()
                accuracy += accuracy_num.item()
            accuracy_list.append(accuracy / len(test_dataset))
    y_list = np.concatenate(y_list)
    y_pre_list = np.concatenate(y_pre_list)

    cm = confusion_matrix(y_list, y_pre_list, normalize='true')
    acc_all = (y_list == y_pre_list).sum()/len(y_list)
    return acc_all, accuracy_list, cm

def test_for_noise(testModel, testDataset, SNR_list: list, batch_size = 64) -> tuple:
    testModel.eval()
    testModel.to(device)
    y_list = []
    y_pre_list = []
    accuracy_list = []
    for snr in tqdm(SNR_list):
        test_dataset = testDataset.get_snr_data(snr)
        test_data = DataLoader(test_dataset, batch_size=batch_size, shuffle=True, drop_last=False)
        with torch.no_grad():
            accuracy = 0.0
            for x, y in test_data:
                x, y = x.to(device), y.to(device)
                denoised_x = x - testModel.denoise(x)
                y_pred = testModel.classifier(denoised_x)
                y_list.append(y.argmax(dim=1).cpu().numpy())
                y_pre_list.append(y_pred.argmax(dim=1).cpu().numpy())
                accuracy_num = (y.argmax(dim=1) == y_pred.argmax(dim=1)).sum()
                accuracy += accuracy_num.item()
            accuracy_list.append(accuracy / len(test_dataset))
    y_list = np.concatenate(y_list)
    y_pre_list = np.concatenate(y_pre_list)

    cm = confusion_matrix(y_list, y_pre_list, normalize='true')
    acc_all = (y_list == y_pre_list).sum()/len(y_list)
    return acc_all, accuracy_list, cm

def visualization(snr_list, acc_list, mode, cm):
    plt.figure(figsize=(10, 6))
    ax = plt.gca()

    # 绘制正确率曲线
    plt.plot(snr_list, acc_list,
             marker='o',
             markersize=8,
             linewidth=2,
             color='#2c7bb6',
             markerfacecolor='#d7191c',
             markeredgecolor='white',
             markeredgewidth=1.5)


    plt.fill_between(snr_list, acc_list,
                     color='#abd9e9',
                     alpha=0.4)


    plt.xlim([-10, 20])
    plt.ylim([0.0, 1.01])
    plt.xticks(np.arange(-10, 21, 5), fontsize=12)
    plt.yticks(np.arange(0, 1.05, 0.1), fontsize=12)
    plt.xlabel('SNR (dB)', fontsize=14, labelpad=10)
    plt.ylabel('Accuracy', fontsize=14, labelpad=10)
    ax.grid(True, which='major', linestyle='--', linewidth=0.5, color='gray')
    ax.set_axisbelow(True)

    max_acc = max(acc_list)
    max_snr = snr_list[acc_list.index(max_acc)]
    plt.annotate(f'Max Accuracy: {max_acc:.4f}',
                 (max_snr, max_acc),
                 xytext=(-50, -60),
                 textcoords='offset points',
                 arrowprops=dict(arrowstyle="->", connectionstyle="arc3"),
                 fontsize=12)

    avg_acc = np.mean(acc_list)
    plt.axhline(avg_acc, color='#fdae61', linestyle='--', linewidth=1.5)
    plt.text(18, avg_acc + 0.02, f'Avg: {avg_acc:.4f}', ha='right', fontsize=12)

    plt.tight_layout()
    plt.figure()
    ax = sns.heatmap(cm, annot=True, fmt='.2f', cmap="Blues", xticklabels=mode, yticklabels=mode)
    ax.set_xticklabels(ax.get_xticklabels(), rotation=45, fontsize=10)
    ax.set_yticklabels(ax.get_yticklabels(), rotation=0, fontsize=10)
    ax.set_xlabel('Predicted Labels')
    ax.set_ylabel('True Labels')
    plt.tight_layout()
    plt.show()


if __name__=="__main__":
    SNR_LEVEL = list(range(-10, 22, 2))
    dataset = RMLDataset(args.dataset, snr_list=SNR_LEVEL, seed=2025)
    _, TestDataset = dataset.split_data(n_train=0.75)
    mode = dataset.modulation_mode
    acc_list_B = None
    cm_B = None
    if args.type == "resnet":
        model = ResNet(BasicBlock, [2, 2, 2], 10)
        model.load_state_dict(torch.load(args.model, weights_only=True))
        acc_all_B, acc_list_B, cm_B = test(model, TestDataset, SNR_LEVEL)

    if args.type == "unet-resnet":
        classifier_model = ResNet(BasicBlock, [2, 2, 2], num_classes=10)
        denoise_model = Unet()
        model = CascadeModel(denoise_model, classifier_model)
        model.load_state_dict(torch.load(args.model, weights_only=True))
        acc_all_B, acc_list_B, cm_B = test(model, TestDataset, SNR_LEVEL)


    if args.type == "unet-noise":
        classifier_model = ResNet(BasicBlock, [2, 2, 2], num_classes=10)
        denoise_model = Unet()
        model = CascadeModel(denoise_model, classifier_model)
        model.modelA.load_state_dict(torch.load(args.model, weights_only=True))
        model.modelB.load_state_dict(torch.load(args.cls_model, weights_only=True))
        acc_all_B, acc_list_B, cm_B = test_for_noise(model, TestDataset, SNR_LEVEL)


    if args.type == "unet-signal":
        classifier_model = ResNet(BasicBlock, [2, 2, 2], num_classes=10)
        denoise_model = Unet()
        model = CascadeModel(denoise_model, classifier_model)
        model.modelA.load_state_dict(torch.load(args.model, weights_only=True))
        model.modelB.load_state_dict(torch.load(args.cls_model, weights_only=True))
        acc_all_B, acc_list_B, cm_B = test(model, TestDataset, SNR_LEVEL)
    if acc_list_B is not None and cm_B is not None:
        visualization(SNR_LEVEL, acc_list_B, mode, cm_B)