Transfer-testing/modelTesting.py at master · zabor2432/Transfer-testing · 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
import os
import tensorflow as tf
from keras.models import load_model
import json
import matplotlib.pyplot as plt
import numpy as np

BATCH_SIZE = 32
MODELS_PATH = '/home/macierz/s181655/projektBadawczy/Transfer-testing/Books/models'
TEST_PATH = '/home/macierz/s181655/projektBadawczy/Transfer-testing/Books'
METRICS_PATH = '/home/macierz/s177788/projektBadawczy/Transfer-testing/modelTesting'
CONSIDERED_MODELS = ['AB_1_10', 'AB_1_25', 'AB_1_50', 'AB_1_100', 'AB_3_10', 'AB_3_25', 'AB_3_50', 'AB_3_100', 'AB_10_10', 'AB_10_25', 'AB_10_50', 'AB_10_100', 'AB_30_10', 'AB_30_25', 'AB_30_50',
                     'AB_30_100',  'AD_3_10', 'AD_3_25', 'AD_3_50', 'AD_3_100', 'CD_1_10', 'CD_1_25', 'CD_1_50', 'CD_1_100', 'CD_3_10', 'CD_3_25', 'CD_3_50', 'CD_3_100', 'CD_10_10', 'CD_10_25',
                     'CD_10_50', 'CD_10_100', 'GH_1_100', 'GH_3_100','CD_30_10', 'CD_30_25', 'CD_3_50', 'CD_30_100', 'CD_1_50', 'CD_1_100']

def getLoaders(dbPath: str):
    testDir = os.path.join(dbPath, "test")
    testTDir = os.path.join(dbPath, "test_T")
    testTTDir = os.path.join(dbPath, "test_TT")

    test_dataset = tf.keras.utils.image_dataset_from_directory(
        testDir,
        labels="inferred",
        label_mode="categorical",
        batch_size=BATCH_SIZE,
        image_size=(224, 224),
        shuffle=False,
    )

    testT_dataset = tf.keras.utils.image_dataset_from_directory(
        testTDir,
        labels="inferred",
        label_mode="categorical",
        batch_size=BATCH_SIZE,
        image_size=(224, 224),
        shuffle=False,
    )

    testTT_dataset = tf.keras.utils.image_dataset_from_directory(
        testTTDir,
        labels="inferred",
        label_mode="categorical",
        batch_size=BATCH_SIZE,
        image_size=(224, 224),
        shuffle=False,
    )

    test_dataset = test_dataset.prefetch(buffer_size=tf.data.AUTOTUNE)
    testT_dataset = testT_dataset.prefetch(buffer_size=tf.data.AUTOTUNE)
    testTT_dataset = testTT_dataset.prefetch(buffer_size=tf.data.AUTOTUNE)

    return test_dataset, testT_dataset, testTT_dataset

if __name__ == '__main__':

    filesInBooks = os.listdir(TEST_PATH)
    modelsList = os.listdir(MODELS_PATH)
    filteredFiles = [x for x in filesInBooks if x.startswith('TestDataset')]

    for testset in filteredFiles:
        testName = testset.replace('TestDataset', '')
        testName =  testName[:1] + testName[2:] + '_'

        currModels = [x for x in CONSIDERED_MODELS if x.startswith(testName)]
        bigDataVersions = [testset.replace("TestDataset", "Big_Data_") + x.replace(testName, "_") for x in currModels]
        bigDataVersions = [x for x in bigDataVersions if x in modelsList]

        currModels = currModels + bigDataVersions

        test_dataset,testT_dataset,testTT_dataset = getLoaders(os.path.join(TEST_PATH, testset))

        for modelName in currModels:
            if "Big_Data" not in modelName: # comment if everything without BigData
                continue

            modelCurr = os.path.join(MODELS_PATH, modelName)
            avFiles = os.listdir(modelCurr)

            metDic = {}
            testLossArr=[]
            testAccuracyArr =[]
            testTLossArr=[]
            testTAccuracyArr =[]
            testTTLossArr=[]
            testTTAccuracyArr =[]
            epochs = np.arange(1, len(avFiles)+1, 1)

            for file in avFiles:
                if not '.h5' in file:
                    continue
                model = load_model(os.path.join(modelCurr, file))
                metrics = model.evaluate(test_dataset)
                metricsT = model.evaluate(testT_dataset)
                metricsTT = model.evaluate(testTT_dataset)

                metDic[file] = {"testLoss": metrics[0],
                                "testAccuracy": metrics[1],
                                "testTLoss": metricsT[0],
                                "testTAccuracy": metricsT[1],
                                "testTTLoss": metricsTT[0],
                                "testTTAccuracy": metricsTT[1]}

                testLossArr.append(metrics[0])
                testAccuracyArr.append(metrics[1])
                testTLossArr.append(metricsT[0])
                testTAccuracyArr.append(metricsT[1])
                testTTLossArr.append(metricsTT[0])
                testTTAccuracyArr.append(metricsTT[1])


            plt.plot(epochs, testLossArr, color = "deeppink", label = "Test Loss")
            plt.plot(epochs, testTLossArr, color = "darkgreen", label = "TestT Loss")
            plt.plot(epochs, testTTLossArr, label = "TestTT Loss")
            plt.legend()
            plt.title(modelName)
            plt.xlabel("Epochs")
            plt.xticks(np.arange(1, len(avFiles)+1, 1))
            plt.ylabel("Loss")
            plt.savefig(os.path.join(METRICS_PATH, f'{modelName}_LOSS.png'))
            plt.close()

            plt.plot(epochs, testAccuracyArr, color = "deeppink", label = "Test Accuracy")
            plt.plot(epochs, testTAccuracyArr, color = "darkgreen", label = "TestT Accuracy")
            plt.plot(epochs, testTTAccuracyArr, label = "TestTT Accuracy")
            plt.legend()
            plt.title(modelName)
            plt.xlabel("Epochs")
            plt.xticks(np.arange(1, len(avFiles)+1, 1))
            plt.ylabel("Accuracy")
            plt.savefig(os.path.join(METRICS_PATH, f'{modelName}_ACC.png'))
            plt.close()


            with open(os.path.join(METRICS_PATH, f'{modelName}.json'), "w") as f:
                json.dump(metDic, f, indent=2)