-
-
Notifications
You must be signed in to change notification settings - Fork 672
/
Copy pathvgg16.py
97 lines (74 loc) · 4.51 KB
/
vgg16.py
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
#! /usr/bin/env python
# coding=utf-8
#================================================================
# Copyright (C) 2019 * Ltd. All rights reserved.
#
# Editor : VIM
# File name : vgg16.py
# Author : YunYang1994
# Created date: 2019-07-16 10:09:17
# Description :
#
#================================================================
import skimage
import numpy as np
import tensorflow as tf
VGG_MEAN = [103.939, 116.779, 123.68]
# define input layer
input_layer = tf.keras.layers.Input([224, 224, 3])
red, green, blue = tf.split(axis=3, num_or_size_splits=3, value=input_layer)
bgr = tf.concat(axis=3, values=[blue - VGG_MEAN[0], green - VGG_MEAN[1], red - VGG_MEAN[2]])
# Block 1
conv1_1 = tf.keras.layers.Conv2D(filters=64, kernel_size=[3, 3], strides=[1, 1], padding='same',
use_bias=True, activation='relu', name='conv1_1')(bgr)
conv1_2 = tf.keras.layers.Conv2D(filters=64, kernel_size=[3, 3], strides=[1, 1], padding='same',
use_bias=True, activation='relu', name='conv1_2')(conv1_1)
pool1_1 = tf.nn.max_pool(conv1_2, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], padding='SAME', name='pool1_1')
# Block 2
conv2_1 = tf.keras.layers.Conv2D(filters=128, kernel_size=[3, 3], strides=[1, 1], padding='same',
use_bias=True, activation='relu', name='conv2_1')(pool1_1)
conv2_2 = tf.keras.layers.Conv2D(filters=128, kernel_size=[3, 3], strides=[1, 1], padding='same',
use_bias=True, activation='relu', name='conv2_2')(conv2_1)
pool2_1 = tf.nn.max_pool(conv2_2, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], padding='SAME', name='pool2_1')
# Block 3
conv3_1 = tf.keras.layers.Conv2D(filters=256, kernel_size=[3, 3], strides=[1, 1], padding='same',
use_bias=True, activation='relu', name='conv3_1')(pool2_1)
conv3_2 = tf.keras.layers.Conv2D(filters=256, kernel_size=[3, 3], strides=[1, 1], padding='same',
use_bias=True, activation='relu', name='conv3_2')(conv3_1)
conv3_3 = tf.keras.layers.Conv2D(filters=256, kernel_size=[3, 3], strides=[1, 1], padding='same',
use_bias=True, activation='relu', name='conv3_3')(conv3_2)
pool3_1 = tf.nn.max_pool(conv3_3, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], padding='SAME', name='pool3_1')
# Block 4
conv4_1 = tf.keras.layers.Conv2D(filters=512, kernel_size=[3, 3], strides=[1, 1], padding='same',
use_bias=True, activation='relu', name='conv4_1')(pool3_1)
conv4_2 = tf.keras.layers.Conv2D(filters=512, kernel_size=[3, 3], strides=[1, 1], padding='same',
use_bias=True, activation='relu', name='conv4_2')(conv4_1)
conv4_3 = tf.keras.layers.Conv2D(filters=512, kernel_size=[3, 3], strides=[1, 1], padding='same',
use_bias=True, activation='relu', name='conv4_3')(conv4_2)
pool4_1 = tf.nn.max_pool(conv4_3, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], padding='SAME', name='pool4_1')
# Block 4
conv5_1 = tf.keras.layers.Conv2D(filters=512, kernel_size=[3, 3], strides=[1, 1], padding='same',
use_bias=True, activation='relu', name='conv5_1')(pool4_1)
conv5_2 = tf.keras.layers.Conv2D(filters=512, kernel_size=[3, 3], strides=[1, 1], padding='same',
use_bias=True, activation='relu', name='conv5_2')(conv5_1)
conv5_3 = tf.keras.layers.Conv2D(filters=512, kernel_size=[3, 3], strides=[1, 1], padding='same',
use_bias=True, activation='relu', name='conv5_3')(conv5_2)
pool5_1 = tf.nn.max_pool(conv5_3, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], padding='SAME', name='pool5_1')
flatten = tf.keras.layers.Flatten()(pool5_1)
fc6 = tf.keras.layers.Dense(units=4096, use_bias=True, name='fc6', activation='relu')(flatten)
fc7 = tf.keras.layers.Dense(units=4096, use_bias=True, name='fc7', activation='relu')(fc6)
fc8 = tf.keras.layers.Dense(units=1000, use_bias=True, name='fc8', activation=None)(fc7)
prob = tf.nn.softmax(fc8)
# Build model
model = tf.keras.Model(input_layer, prob)
# # Load weighs
# weighs = np.load("./vgg16.npy", encoding='latin1').item()
# for layer_name in weighs.keys():
# layer = model.get_layer(layer_name)
# layer.set_weights(weighs[layer_name])
# # Load image
# image_data = skimage.io.imread("./docs/cat.jpg").astype(np.float32)
# # Load labels
# labels = open("./docs/synset_words.txt", "r").readlines()
# # Print result
# print(labels[np.argmax(model(np.expand_dims(image_data, 0)))])