Fer-Eating/behavior_compare.py at main · famesensor/Fer-Eating · 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
import cv2
import time
import glob
import numpy as np
import matplotlib.pyplot as plt
from datetime import datetime

from behavior.behavior_model import init_model_behavior
from detection.yolo_detector import init_model_person, person_detect
from preparation.preparation import resize_image, normalize_image, load_data_set
from sklearn.metrics import confusion_matrix
from mlxtend.plotting import plot_confusion_matrix
from sklearn.metrics import accuracy_score
from sklearn.metrics import recall_score
from sklearn.metrics import precision_score
from sklearn.metrics import f1_score


if __name__ == "__main__":
    # init value
    img_height, img_weight = 224, 224
    channels = 3  # 3 RGB
    batch_size = 32

    weight_model_vgg16 = "./models/behavior/vgg16/vgg16_behavior.h5"
    weight_model_vgg19 = "./models/behavior/vgg19/vgg19_behavior.h5"
    weight_model_resnet = "./models/behavior/resnet/resnet_behavior.h5"
    weight_model_mobilenet = "./models/behavior/mobilenet/mobilenet_behavior.h5"
    config_person = './models/yolo/yolov4.cfg'
    weight_person = './models/yolo/yolov4.weights'
    include_top = False
    class_num = 2
    activation = "softmax"
    loss = "categorical_crossentropy"
    data_time_path = './dataset/behavior/test_time/*.jpg'
    data_evalute_path = './dataset/behavior/validate/'
    test_datagen_args = dict(
        rescale=1./255
    )

    # init model...
    person_model = init_model_person("./models/yolov4-tensorflow")
    vgg16 = init_model_behavior(weight_path=weight_model_vgg16, types="vgg16", include_top=include_top, img_height=img_height,
                                img_weight=img_weight, channels=channels, class_num=class_num, layer_num=19, activation=activation, loss=loss)

    # vgg19 = init_model_behavior(weight_path=weight_model_vgg19, types="vgg19", include_top=include_top, img_height=img_height,
    #                             img_weight=img_weight, channels=channels, class_num=class_num, layer_num=22, activation=activation, loss=loss)

    # resnet = init_model_behavior(weight_path=weight_model_resnet, types="resnet", include_top=include_top, img_height=img_height,
    #                               img_weight=img_weight, channels=channels, class_num=class_num, layer_num=190, activation=activation, loss=loss)

    # mobile_net = init_model_behavior(weight_path=weight_model_mobilenet, types="mobilenet", include_top=include_top, img_height=img_height,
    #                                   img_weight=img_weight, channels=channels, class_num=class_num, layer_num=154, activation=activation, loss=loss)

    print("[INFO]: compare time model expression...")
    start = datetime.now()
    startTime = time.time()
    print("[INFO]: start time predict vgg19: ", start)
    images = glob.glob(data_time_path)
    res_predict = []
    test_data = []
    index = 0
    for image_path in images:
        image = cv2.imread(image_path)

        # face dectection
        person_res = person_detect(image, person_model)
        if not person_res.any():
            continue

        # preparation data for expression model
        person_res = resize_image(
            image=person_res, size_image=(img_height, img_weight))
        person_res = normalize_image(image=person_res)
        person_res = np.expand_dims(person_res, axis=0)

        res_predict.append(vgg16.predict(person_res))
        # res_predict.append(vgg19.predict(person_res))
        # res_predict.append(resnet.predict(person_res))
        # res_predict.append(mobile_net.predict(person_res))
        print(image_path)

    # Time elapsed
    end = time.time()
    duration = datetime.now() - start
    print("[INFO]: time predict vgg19 : ", duration)
    seconds = end - startTime
    fps = len(res_predict)/seconds
    print(f"[INFO]: images : {len(res_predict)}")
    print("[INFO]: Estimated frames per second : {0}".format(fps))

    # # print("[INFO]: compare evaluate model expression...")
    # test_data = load_data_set(
    #     test_datagen_args, data_evalute_path, (img_height, img_weight), batch_size)

    # # vgg16_evaluate[0] is loss, vgg16_evaluate[1] is accreency
    # # vgg16_evaluate = vgg16.evaluate(test_data)

    # vgg19_evaluate = vgg19.evaluate(test_data)

    # # resnet_evaluate = resnet.evaluate(test_data)

    # # mobile_evaluate = mobile_net.evaluate(test_data)

    # # Evaluation on test dataset
    # print("test loss, test acc:", vgg19_evaluate)

    # # Plot the confusion matrix
    # test_logits = vgg19.predict(test_data)
    # rounded_labels = np.argmax(test_logits, axis=1)

    # cm = confusion_matrix(test_data.classes, np.round(rounded_labels))
    # class_names = ['eat', 'dont eat']
    # # Plot Non-Normalized
    # plot_confusion_matrix(cm, class_names=class_names)
    # # Plot Normalized
    # plot_confusion_matrix(cm, show_absolute=False,
    #                       show_normed=True, class_names=class_names)
    # plt.show()

    # print("Accuracy Score :", accuracy_score(test_data.labels, rounded_labels))

    # print("Recall Score : ", recall_score(
    #     test_data.labels, rounded_labels, average='macro'))

    # print("Precision Score : ", precision_score(
    #     test_data.labels, rounded_labels, average='macro'))

    # print("F1 Score : ", f1_score(
    #     test_data.labels, rounded_labels, average='macro'))