Mark-micropython/local_training.py at master · TinkerGen/Mark-micropython · 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
from camera import snapshot, global_value, rows
import image, lcd, time
import KPU as kpu
import gc
import ujson

dict_add = {}
class OnDeviceTraining(object):
    def __init__(self, categoryname=[], samplesnumber=15, bool=False):
        self.row = global_value.row
        global_value.row = global_value.row + 1
        try:
            del self.model
        except Exception:
            pass
        try:
            del self.classifier
        except Exception:
            pass

        self.classnumber = len(categoryname)
        self.samplesnumber = samplesnumber
        self.categoryname = categoryname
        self.img_copy = None
        self.cap_num = 0
        self.train_status = 0
        self.class_list=[]
        self.percent = 0
        self.image_class = 0

        if bool:
            self.model = kpu.load(0x514000)
            self.classifier = kpu.classifier(self.model, self.classnumber, self.samplesnumber)
        else:
            pass

    def classify_image(self, string):
        img = snapshot()
        self.img_copy = img.resize(224,224)
        a = self.img_copy.pix_to_ai()
        a = img.draw_string(0, rows[self.row], string, color=(255,255,255), scale=1.5, mono_space=False)
        a = lcd.display(img)
        del(img)
        gc.collect()

    # parameter 1-5
    def record_seed_sample(self, parameter=''):
        self.classify_image(("Image for %s is taken" % parameter))
        if self.cap_num < self.classnumber:
            class_index = self.categoryname.index(parameter)
            index = self.classifier.add_class_img(self.img_copy)
            self.cap_num += 1
            self.class_list.append(class_index)

    def record_samples_training(self):
        self.classify_image(("Samples taken: %d" % (self.cap_num-3)))
        if self.cap_num < self.classnumber + self.samplesnumber:
            index = self.classifier.add_sample_img(self.img_copy)
            self.cap_num += 1

        if self.train_status == 0:
            if self.cap_num >= self.classnumber + self.samplesnumber:
                print("start train")
                img = snapshot()
                a = img.draw_string(lcd.width()//2-68,lcd.height()//2-4, "Training...", color=(0,255,0), scale=3, mono_space=False)
                lcd.display(img)
                del(img)
                #del(self.img_copy)
                gc.collect()
                time.sleep(2)
                self.classifier.train()
                print("train end")
                self.train_status = 1

    def class_predict(self, THRESHOLD=50):
        img = snapshot()
        self.img_copy = img.resize(224,224)
        #kpu.memtest()
        a = self.img_copy.pix_to_ai()
        res_index = -1
        try:
            res_index, min_dist = self.classifier.predict(self.img_copy)
            self.percent = round(min_dist, 0)
            #print(" percent:",self.percent)
            #print("res_index: ", res_index)
            #print("{:.2f}".format(min_dist))
        except Exception as e:
            print("predict err:", e)
        if res_index >= 0 and self.percent >= THRESHOLD :
            self.image_class = self.categoryname[self.class_list[res_index]]
            #print("predict result:", self.image_class)

        a = lcd.display(img)
        del(img)
        #del(img_copy)
        gc.collect()

    def save_model_file(self, filename=''):
        self.classifier.save(filename)
        #print(filename.split('.')[0]+".names")
        f = open(filename.split('.')[0]+".names",'w')
        #print(str(self.class_list))
        f.write(str(self.class_list))
        f.close()

    def load_model_file(self, filename=''):
        self.model = kpu.load(0x514000)
        classifier = kpu.classifier.load(self.model, filename)
        self.classifier = classifier[0]
        f = open(filename.split('.')[0]+'.names','r')
        self.class_list = ujson.loads(f.read())
        #print(self.class_list)
        f.close()

    def get_classification_result(self, percent):
        threshold = percent/100
        self.class_predict(threshold)
        #gc.collect()
        if self.percent >= threshold:
            return self.image_class
        else:
            return -1

    def is_class(self, _class, percent):
        threshold = percent/100
        self.class_predict(threshold)
        #gc.collect()
        if self.image_class == _class and self.percent >= threshold:
            return True
        else:
            return False


"""
filename = "/sd/123.tg"
categoryname = ['a', 'b', 'c']
samplesnumber = 15 #15-25

modeltrain = OnDeviceTraining(categoryname, samplesnumber, 1)
time.sleep(4)
modeltrain.record_seed_sample(categoryname[2])
time.sleep(4)
modeltrain.record_seed_sample(categoryname[1])
time.sleep(4)
modeltrain.record_seed_sample(categoryname[0])
time.sleep(4)

for a in range(15):
    modeltrain.record_samples_training()
    time.sleep(3)


modeltrain.save_model_file(filename)
#modeltrain.load_model_file(filename)

while True:
    pass
    #modeltrain.get_classification_result(50)
    modeltrain.is_class('a', 20)

"""