text
stringlengths
1
93.6k
# get filename list
tfrecord_filename = tf.gfile.Glob(self.tfrecord_file + '*%s*' % self.name)
print('tfrecord test filename', tfrecord_filename)
# The file name list generator
filename_queue = tf.train.string_input_producer(tfrecord_filename, num_epochs=None, shuffle=False)
# get tensor of image/label
image, label = read_tfrecord_and_decode_into_image_label_pair_tensors(filename_queue,
self.actual_image_size)
#image = channels_image_standardization(image)
image = image_standardization(image)
image_batch, label_batch = tf.train.batch([image, label],
batch_size=self.batch_size,
capacity=self.capacity)
# num_threads=self.num_threads)
image_batch = tf.image.resize_bilinear(image_batch, size=[self.actual_image_size, self.actual_image_size])
return image_batch, label_batch
def read_tfrecord_and_decode_into_image_label_pair_tensors(tfrecord_filenames_queue, size):
"""Return label/image tensors that are created by reading tfrecord file.
The function accepts tfrecord filenames queue as an input which is usually
can be created using tf.train.string_input_producer() where filename
is specified with desired number of epochs. This function takes queue
produced by aforemention tf.train.string_input_producer() and defines
tensors converted from raw binary representations into
reshaped label/image tensors.
Parameters
----------
tfrecord_filenames_queue : tfrecord filename queue
String queue object from tf.train.string_input_producer()
Returns
-------
image, label : tuple of tf.int32 (image, label)
Tuple of label/image tensors
"""
reader = tf.TFRecordReader()
_, serialized_example = reader.read(tfrecord_filenames_queue)
features = tf.parse_single_example(
serialized_example,
features={
'image/height': tf.FixedLenFeature([], tf.int64),
'image/width': tf.FixedLenFeature([], tf.int64),
'image/depth': tf.FixedLenFeature([], tf.int64),
'image/encoded': tf.FixedLenFeature([], tf.string),
'image/class/label': tf.FixedLenFeature([], tf.int64),
# 'image': tf.FixedLenFeature([], tf.string)
})
image = tf.decode_raw(features['image/encoded'], tf.uint8)
label = tf.cast(features['image/class/label'], tf.int64)
height = tf.cast(features['image/height'], tf.int64)
width = tf.cast(features['image/width'], tf.int64)
depth = tf.cast(features['image/depth'], tf.int64)
image = tf.reshape(image, [128,128,3]) #height,width,depth
print(image)
image = tf.to_float(image)
return image, label
def image_standardization(image):
out_image = image/255.0
#out_image = image/127.5 - 1.0
'''out_image = image
#out_image = tf.reverse(image, axis=[-1])
rgb = tf.cast(np.array([[[123, 117, 104]]]), tf.float32)
out_image = out_image - rgb
out_image = out_image/127.5'''
return out_image
def dataaugmentation(images):
images = tf.image.random_flip_up_down(images)
images = tf.image.random_flip_left_right(images)
#images = tf.image.random_brightness(images, max_delta=0.3)
#images = tf.image.random_contrast(images, 0.8, 1.2)
#images = tf.image.random_saturation(images, 0.7, 1.3)
return images
# <FILESEP>
"""
Copyright (c) Microsoft Corporation.
Licensed under the MIT license.
UNITER pre-training
"""
import argparse
from collections import defaultdict
import json
import os
from os.path import exists, join
from time import time
import torch
from torch.utils.data import DataLoader
from torch.nn import functional as F
from torch.nn.utils import clip_grad_norm_