VGG16

VGG16 (presented by University of Oxford)
implemented with TensorFlow

Dependencies

python3.6

• numpy
• skimage
• TensorFlow
• matplotlib

Usage

1. Import required modules

import tensorflow as tf
from util.util import *
from model.vgg16 import *

2. Load test-image

img = load_image('./test.jpg')
img = img.reshape((1, 224, 224, 3))

In this example, load single-image.
If you attempt to batch-process, load some images and concatenate them. Then, modify img-shape e.g.,

img = img.reshape((batch_size, 224, 224, 3))

3. Start Session

with tf.Session() as sess:
    image = tf.placeholder(shape=[batch_size, 224, 224, 3], dtype=tf.float32)
    feed_dict = {image: img}
    vgg = Vgg16()
    vgg.build(image)
    sess.run(tf.global_variables_initializer())

    prob = sess.run(vgg.net, feed_dict=feed_dict)
    print(prob)

Test Training

$ python cifar.py

Present Circumstances

Finished cifar-10 learning.

Complete:

• Learning cifar10
• saving and restoring parameters

If I have overlooked something, please tell me.

Welcome PullRequest or E-mail