Port of @MarvinTeichmann tensorflow-fnc model

Combines the vgg example in tflearn with the FCN example from
tensorflow-fcn

https://github.com/MarvinTeichmann/tensorflow-fcn/blob/master/test_fcn32_vgg.py
https://github.com/tflearn/tflearn/blob/master/examples/images/vgg_network.py

examples/images/vgg_fcn_network.py

@@ -0,0 +1,100 @@
+# -*- coding: utf-8 -*-
+
+""" Very Deep Convolutional Networks for Large-Scale Visual Recognition.
+
+VGG 16-layers convolutional with semantic segmentation
+
+References:
+    Fully Convolutional Networks for Semantic Segmentation
+    Jonathan Long*, Evan Shelhamer*, and Trevor Darrell. CVPR 2015.
+
+Links:
+    https://arxiv.org/abs/1605.06211
+
+"""
+
+from __future__ import division, print_function, absolute_import
+
+import tensorflow as tf
+import tflearn
+from tflearn.layers.core import input_data, dropout, fully_connected
+from tflearn.layers.conv import conv_2d, max_pool_2d, upscore_layer, score_layer
+from tflearn.layers.estimator import regression
+
+# Data loading and preprocessing
+# import tflearn.datasets.oxflower17 as oxflower17
+# X, Y = oxflower17.load_data(one_hot=True)
+train = False
+num_classes = 17
+
+# Building 'VGG Network'
+network = input_data(shape=[None, 224, 224, 3])
+
+# Conv 1
+network = conv_2d(network, 64, 3, activation='relu')
+network = conv_2d(network, 64, 3, activation='relu')
+network = max_pool_2d(network, 2, strides=2)
+
+# Conv 2
+network = conv_2d(network, 128, 3, activation='relu')
+network = conv_2d(network, 128, 3, activation='relu')
+network = max_pool_2d(network, 2, strides=2)
+
+# Conv 3
+network = conv_2d(network, 256, 3, activation='relu')
+network = conv_2d(network, 256, 3, activation='relu')
+network = conv_2d(network, 256, 3, activation='relu')
+pool3 = max_pool_2d(network, 2, strides=2)
+
+# Conv 4
+network = conv_2d(pool3, 512, 3, activation='relu')
+network = conv_2d(network, 512, 3, activation='relu')
+network = conv_2d(network, 512, 3, activation='relu')
+pool4 = max_pool_2d(network, 2, strides=2)
+
+# Conv 5
+network = conv_2d(pool4, 512, 3, activation='relu')
+network = conv_2d(network, 512, 3, activation='relu')
+network = conv_2d(network, 512, 3, activation='relu')
+network = max_pool_2d(network, 2, strides=2)
+
+# FC 6
+network = conv_2d(network, 4096, 7, activation='relu')
+network = dropout(network, 0.5)
+
+# FC 7
+network = conv_2d(network, 1000, 1, activation='relu')
+network = dropout(network, 0.5)
+
+pred = tf.argmax(network, dimension=3)
+
+upscore2 = upscore_layer(network,
+                         shape=tf.shape(pool4),
+                         num_classes=num_classes,
+                         kernel_size=4, strides=2, name='upscore2')
+
+score_pool4 = score_layer(pool4, num_classes=num_classes, name='score_pool4')
+fuse_pool4 = tf.add(upscore2, score_pool4)
+
+upscore4 = upscore_layer(fuse_pool4, num_classes=num_classes,
+                         kernel_size=4, strides=2, name='upscore4')
+
+score_pool3 = score_layer(pool3, num_classes=num_classes, name='score_pool3')
+fuse_pool3 = tf.add(upscore4, score_pool3)
+
+upscore32 = upscore_layer(fuse_pool3, num_classes=num_classes,
+                          kernel_size=16, strides=8, name='upscore32')
+
+
+network = regression(network, optimizer='adam',
+                     loss='categorical_crossentropy',
+                     learning_rate=0.001)
+
+# Training
+model = tflearn.DNN(network, checkpoint_path='model_vgg',
+                    max_checkpoints=1, tensorboard_verbose=0)
+
+if train:
+    model.fit(X, Y, n_epoch=500, shuffle=True,
+              show_metric=True, batch_size=32, snapshot_step=500,
+              snapshot_epoch=False, run_id='vgg_oxflowers17')