def setup_record(self):
filename_queue = tf.train.string_input_producer(
[self.record], num_epochs=10)
self.image, self.annotation = read_tfrecord_and_decode_into_image_annotation_pair_tensors(filename_queue)
self.resized_image, resized_annotation = scale_randomly_image_with_annotation_with_fixed_size_output(self.image, self.annotation, (self.size, self.size))
self.resized_annotation = tf.squeeze(resized_annotation)
def setup_val(self, tfname):
self.restore = glob(os.path.join(self.checkpoint8, "FCN__*", "*.data*" ))[0].split(".data")[0]
filename_queue = tf.train.string_input_producer(
[tfname], num_epochs=10)
self.image_queue, self.annotation_queue = read_tfrecord_and_decode_into_image_annotation_pair_tensors(filename_queue)
self.image = tf.placeholder_with_default(self.image, shape=[None,
None,
3])
self.annotation = tf.placeholder_with_default(self.annotation_queue, shape=[None,
None,
1])
self.resized_image, resized_annotation = scale_randomly_image_with_annotation_with_fixed_size_output(self.image, self.annotation, (self.size, self.size))
self.resized_annotation = tf.squeeze(resized_annotation)
image_batch_tensor = tf.expand_dims(self.image, axis=0)
annotation_batch_tensor = tf.expand_dims(self.annotation, axis=0)
# Be careful: after adaptation, network returns final labels
# and not logits
FCN_8s_bis = adapt_network_for_any_size_input(FCN_8s, 32)
self.pred, fcn_16s_variables_mapping = FCN_8s_bis(image_batch_tensor=image_batch_tensor,
number_of_classes=self.num_labels,
is_training=False)
self.prob = [h for h in [s for s in [t for t in self.pred.op.inputs][0].op.inputs][0].op.inputs][0]
initializer = tf.local_variables_initializer()
self.saver = tf.train.Saver()
with tf.Session() as sess:
sess.run(initializer)
self.saver.restore(sess, self.restore)
def setup_record(self):
"""
Setup record reading.
"""
filename_queue = tf.train.string_input_producer([self.record],
num_epochs=10)
self.image, self.annotation = read_tfrecord_and_decode_into_image_annotation_pair_tensors(
filename_queue)
self.resized_image, resized_annotation = scale_randomly_image_with_annotation_with_fixed_size_output(
self.image, self.annotation, (self.size, self.size))
self.resized_annotation = tf.squeeze(resized_annotation)
def setup_val(self, tfname):
"""
Setups the model in case we need to validate.
"""
self.restore = glob(os.path.join(self.checkpoint8, "FCN__*",
"*.data*"))[0].split(".data")[0]
filename_queue = tf.train.string_input_producer([tfname],
num_epochs=10)
self.image_queue, self.annotation_queue = read_tfrecord_and_decode_into_image_annotation_pair_tensors(
filename_queue)
self.image = tf.placeholder_with_default(self.image,
shape=[None, None, 3])
self.annotation = tf.placeholder_with_default(self.annotation_queue,
shape=[None, None, 1])
self.resized_image, resized_annotation = scale_randomly_image_with_annotation_with_fixed_size_output(
self.image, self.annotation, (self.size, self.size))
self.resized_annotation = tf.squeeze(resized_annotation)
image_batch_tensor = tf.expand_dims(self.image, axis=0)
annotation_batch_tensor = tf.expand_dims(self.annotation, axis=0)
# Be careful: after adaptation, network returns final labels
# and not logits
FCN_8s_bis = adapt_network_for_any_size_input(FCN_8s, 32)
self.pred, fcn_16s_variables_mapping = FCN_8s_bis(
image_batch_tensor=image_batch_tensor,
number_of_classes=self.num_labels,
is_training=False)
self.prob = [
h for h in
[s
for s in [t
for t in self.pred.op.inputs][0].op.inputs][0].op.inputs
][0]
initializer = tf.local_variables_initializer()
self.saver = tf.train.Saver()
with tf.Session() as sess:
sess.run(initializer)
self.saver.restore(sess, self.restore)
from matplotlib import pyplot as plt
from tf_image_segmentation.utils.pascal_voc import pascal_segmentation_lut
from tf_image_segmentation.utils.tf_records import read_tfrecord_and_decode_into_image_annotation_pair_tensors
from tf_image_segmentation.utils.inference import adapt_network_for_any_size_input
from tf_image_segmentation.utils.visualization import visualize_segmentation_adaptive
pascal_voc_lut = pascal_segmentation_lut()
tfrecord_filename = 'pascal_augmented_val.tfrecords'
number_of_classes = 21
filename_queue = tf.train.string_input_producer([tfrecord_filename],
num_epochs=1)
image, annotation = read_tfrecord_and_decode_into_image_annotation_pair_tensors(
filename_queue)
# Fake batch for image and annotation by adding
# leading empty axis.
image_batch_tensor = tf.expand_dims(image, axis=0)
annotation_batch_tensor = tf.expand_dims(annotation, axis=0)
# Be careful: after adaptation, network returns final labels
# and not logits
FCN_8s = adapt_network_for_any_size_input(FCN_8s, 32)
pred, fcn_16s_variables_mapping = FCN_8s(image_batch_tensor=image_batch_tensor,
number_of_classes=number_of_classes,
is_training=False)
# Take away the masked out values from evaluation
def read_photo_annotation_depth(batch_size=1):
def convert2float(image):
""" Transfrom from int image ([0,255]) to float tensor ([-1.,1.])
"""
image = tf.image.convert_image_dtype(image, dtype=tf.float32)
return (image / 127.5) - 1.0
image_train_size = [256, 256]
number_of_classes = 21
# tfrecord_filename = '/home/yaxing/DATA/scene_net_augmented_train_200.tfrecords'
tfrecord_filename = '/home/yaxing/data/scene5M/scene_net_augmented_train_16k.tfrecords'
pascal_voc_lut = {
0: 'Unknown',
1: 'Bed',
2: 'Books',
3: 'Ceiling',
4: 'Chair',
5: 'Floor',
6: 'Furniture',
7: 'Objects',
8: 'Picture',
9: 'Sofa',
10: 'Table',
11: 'TV',
12: 'Wall',
13: 'Window'
}
class_labels = pascal_voc_lut.keys()
filename_queue = tf.train.string_input_producer([tfrecord_filename],
num_epochs=50000)
image1, image2, annotation1, depth2, annotation2, depth1 = read_tfrecord_and_decode_into_image_annotation_pair_tensors(
filename_queue)
# Various data augmentation stages
image1, annotation1, depth2, image2, annotation2, depth1 = flip_randomly_left_right_image_with_annotation(
image1, annotation1, depth2, image2, annotation2, depth1)
# image = distort_randomly_image_color(image)
resized_image1, resized_annotation1, resized_depth2, resized_image2, resized_annotation2, resized_depth1 = scale_randomly_image_with_annotation_with_fixed_size_output(
image1,
annotation1,
depth2,
image_train_size,
img_tensor2=image2,
annotation_tensor2=annotation2,
depth_tensor1=depth1)
resized_annotation1 = tf.squeeze(resized_annotation1)
resized_annotation2 = tf.squeeze(resized_annotation2)
image_batch1, annotation_batch1, depth_batch2, image_batch2, annotation_batch2, depth_batch1 = tf.train.shuffle_batch(
[
resized_image1, resized_annotation1, resized_depth2,
resized_image2, resized_annotation2, resized_depth1
],
batch_size=batch_size,
capacity=3000,
num_threads=2,
min_after_dequeue=1000)
return tf.to_float(image_batch1) / 127.5 - 1., annotation_batch1, (
tf.exp(depth_batch1) - 1.
) / 1000., tf.to_float(image_batch2) / 127.5 - 1., annotation_batch2, (
tf.exp(depth_batch2) - 1.) / 1000.
restoremodel = options.checkpoint
lr = restoremodel.split('__')[1]
slim = tf.contrib.slim
tfrecord_filename = options.tf_records
number_of_classes = options.labels
filename_queue = tf.train.string_input_producer(
[tfrecord_filename], num_epochs=1)
image, annotation = read_tfrecord_and_decode_into_image_annotation_pair_tensors(filename_queue)
# Fake batch for image and annotation by adding
# leading empty axis.
image_batch_tensor = tf.expand_dims(image, axis=0)
annotation_batch_tensor = tf.expand_dims(annotation, axis=0)
# Be careful: after adaptation, network returns final labels
# and not logits
FCN_8s = adapt_network_for_any_size_input(FCN_8s, 32)
pred, fcn_16s_variables_mapping = FCN_8s(image_batch_tensor=image_batch_tensor,
number_of_classes=number_of_classes,
is_training=False)
# Take away the masked out values from evaluation
