GM-TCNet/Common_Model.py at main · aascode/GM-TCNet · 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
import sys
from typing import Tuple
from Config import Config
import numpy as np
from sklearn.metrics import accuracy_score
import prettytable as pt

class Common_Model(object):

    def __init__(self, save_path: str = '', name: str = 'Not Specified'):
        self.model = None
        self.trained = False # 模型是否已训练

    '''
    train(): 在给定训练集上训练模型

    输入:
        x_train: 训练集样本
        y_train: 训练集标签
        x_val: 测试集样本
        y_val: 测试集标签

    '''
    def train(self, x_train, y_train, x_val, y_val):
        raise NotImplementedError()

    '''
    predict(): 识别音频的情感

    输入:
        samples: 需要识别的音频特征

    输出:
        list: 识别结果（标签）的list
    '''
    def predict(self, samples):
        raise NotImplementedError()


    '''
    predict_proba(): 音频的情感的置信概率

    输入:
        samples: 需要识别的音频特征

    输出:
        list: 每种情感的概率
    '''
    def predict_proba(self, samples):
        if not self.trained:
            sys.stderr.write("No Model.")
            sys.exit(-1)
        return self.model.predict_proba(samples)

    '''
    save_model(): 将模型以 model_name 命名存储在 /Models 目录下
    '''
    def save_model(self, model_name: str):
        raise NotImplementedError()

    '''
    evaluate(): 在测试集上评估模型，输出准确率

    输入:
        x_test: 样本
        y_test: 标签
    '''
    def evaluate(self, x_test, y_test):
        tb = pt.PrettyTable()
        temp = ["###"]
        print(temp)
        for item in Config.CLASS_LABELS:
            temp.append(item)
        temp.append("All")
        temp.append("Correct")
        temp.append("Accuracy")
        tb.field_names = temp
        predictions = self.predict(x_test)
        y_test = np.argmax(y_test,axis=1)
        num = len(y_test)
        #情感分类统计
        emotion_num = np.zeros((20, 20), dtype=np.int)
        print(y_test)
        print(predictions)
        for i in range(num):
             emotion_num[y_test[i]][10] += 1#表示总的数量
             emotion_num[y_test[i]][predictions[i]] += 1#表示各个情感的数量
             if y_test[i]==predictions[i]:
                 emotion_num[y_test[i]][11]+=1#表示标记正确的数量

        for i in range(len(Config.CLASS_LABELS)):
            print(i,'类acc:',emotion_num[i][11]/emotion_num[i][10])

        print('Accuracy:%.3f\n' % accuracy_score(y_pred = predictions, y_true = y_test))


        for i in range(len(Config.CLASS_LABELS)):
            temp = []
            temp.append(Config.CLASS_LABELS[i])
            for j in range(len(Config.CLASS_LABELS)):
                temp.append(emotion_num[i][j])
            temp.append(emotion_num[i][10])
            temp.append(emotion_num[i][11])
            temp.append(emotion_num[i][11]/emotion_num[i][10])
            tb.add_row(temp)

        print(tb)
        '''
        predictions = self.predict(x_test)
        score = self.model.score(x_test, y_test)
        print("True Lable: ", y_test)
        print("Predict Lable: ", predictions)
        print("Score: ", score)
        '''