-
Notifications
You must be signed in to change notification settings - Fork 0
/
Prediction.jl
52 lines (41 loc) · 2.14 KB
/
Prediction.jl
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
py"""
from __future__ import absolute_import, division, print_function, unicode_literals
import tensorflow as tf
from tensorflow.keras import layers, models
from efficientnet.tfkeras import EfficientNetB5
from tensorflow.keras import Model
from tensorflow.keras.optimizers import Adam
import numpy as np
gpus = tf.config.experimental.list_physical_devices('GPU')
if gpus:
for gpu in gpus:
tf.config.experimental.set_memory_growth(gpu, True)
logical_gpus = tf.config.experimental.list_logical_devices('GPU')
print(len(gpus), "Physical GPUs,", len(logical_gpus), "Logical GPUs")
def predict( test_images , weights_name = 'Weights_giga_B5.h5' ):
conv_base = EfficientNetB0(weights=None, include_top=False, input_shape=(50, 50, 3))
dropout_rate = 0.35
reg = 0.0003
model = models.Sequential()
model.add(conv_base)
# model.add(layers.GlobalAvgPool2D())
model.add(layers.Dense(1024, activation="selu", kernel_regularizer=regularizers.l1(reg), activity_regularizer=regularizers.l1(reg) ) )
model.add(tf.keras.layers.BatchNormalization())
model.add(layers.Dropout(dropout_rate))
model.add(layers.Flatten())
model.add(layers.Dense(512, activation="selu", kernel_regularizer=regularizers.l1(reg), activity_regularizer=regularizers.l1(reg) ) )
model.add(layers.Dropout(dropout_rate))
model.add(tf.keras.layers.BatchNormalization())
model.add(layers.Dense(256, activation="selu", kernel_regularizer=regularizers.l1(reg), activity_regularizer=regularizers.l1(reg) ) )
model.add(layers.Dropout(dropout_rate))
model.add(tf.keras.layers.BatchNormalization())
model.add(layers.Dense(128, activation="selu", kernel_regularizer=regularizers.l1(reg), activity_regularizer=regularizers.l1(reg) ) )
model.add(layers.Dropout(dropout_rate))
model.add(tf.keras.layers.BatchNormalization())
model.add(layers.Dense(1, activation="sigmoid", name="fc_out"))
model.compile(optimizer=Adam(lr=0.0000134),loss = 'binary_crossentropy', metrics=['accuracy'])
model.summary()
model.load_weights(weights_name)
eval_predict = model.predict(test_images)
return np.array(eval_predict)
"""