Skip to content

GitLab

Commit 70b894bd authored by derekjchow's avatar derekjchow Committed by Sergio Guadarrama
Browse files

Add imagenet scripts. (#2185)

parent 6e93d249
Hide whitespace changes
Inline Side-by-side
Showing with 74150 additions and 1 deletion
slim/BUILD
View file @ 70b894bd
......@@ -17,6 +17,28 @@ py_library(
],
)
sh_binary(
name = "download_and_convert_imagenet",
srcs = ["datasets/download_and_convert_imagenet.sh"],
data = [
"datasets/download_imagenet.sh",
"datasets/imagenet_2012_validation_synset_labels.txt",
"datasets/imagenet_lsvrc_2015_synsets.txt",
"datasets/imagenet_metadata.txt",
"datasets/preprocess_imagenet_validation_data.py",
"datasets/process_bounding_boxes.py",
":build_imagenet_data",
],
)
py_binary(
name = "build_imagenet_data",
srcs = ["datasets/build_imagenet_data.py"],
deps = [
"//tensorflow",
],
)
py_library(
name = "download_and_convert_cifar10",
srcs = ["datasets/download_and_convert_cifar10.py"],
......
slim/README.md
View file @ 70b894bd
......@@ -163,7 +163,56 @@ dataset = flowers.get_split('validation', DATA_DIR)
provider = slim.dataset_data_provider.DatasetDataProvider(dataset)
[image, label] = provider.get(['image', 'label'])
```
## An automated script for processing ImageNet data.
Training a model with the ImageNet dataset is a common request. To facilitate
working with the ImageNet dataset, we provide an automated script for
downloading and processing the ImageNet dataset into the native TFRecord
format.
The TFRecord format consists of a set of sharded files where each entry is a serialized `tf.Example` proto. Each `tf.Example` proto contains the ImageNet image (JPEG encoded) as well as metadata such as label and bounding box information.
We provide a single [script](datasets/download_and_preprocess_imagenet.sh) for
downloading and converting ImageNet data to TFRecord format. Downloading and
preprocessing the data may take several hours (up to half a day) depending on
your network and computer speed. Please be patient.
To begin, you will need to sign up for an account with [ImageNet]
(http://image-net.org) to gain access to the data. Look for the sign up page,
create an account and request an access key to download the data.
After you have `USERNAME` and `PASSWORD`, you are ready to run our script. Make
sure that your hard disk has at least 500 GB of free space for downloading and
storing the data. Here we select `DATA_DIR=$HOME/imagenet-data` as such a
location but feel free to edit accordingly.
When you run the below script, please enter *USERNAME* and *PASSWORD* when
prompted. This will occur at the very beginning. Once these values are entered,
you will not need to interact with the script again.
```shell
# location of where to place the ImageNet data
DATA_DIR=$HOME/imagenet-data
# build the preprocessing script.
bazel build slim/download_and_preprocess_imagenet
# run it
bazel-bin/slim/download_and_preprocess_imagenet "${DATA_DIR}"
```
The final line of the output script should read:
```shell
2016-02-17 14:30:17.287989: Finished writing all 1281167 images in data set.
```
When the script finishes you will find 1024 and 128 training and validation
files in the `DATA_DIR`. The files will match the patterns `train-????-of-1024`
and `validation-?????-of-00128`, respectively.
[Congratulations!](https://www.youtube.com/watch?v=9bZkp7q19f0) You are now
ready to train or evaluate with the ImageNet data set.
# Pre-trained Models
<a id='Pretrained'></a>
......
slim/datasets/build_imagenet_data.py 0 → 100644
View file @ 70b894bd
# Copyright 2016 Google Inc. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# ==============================================================================
"""Converts ImageNet data to TFRecords file format with Example protos.
The raw ImageNet data set is expected to reside in JPEG files located in the
following directory structure.
data_dir/n01440764/ILSVRC2012_val_00000293.JPEG
data_dir/n01440764/ILSVRC2012_val_00000543.JPEG
...
where 'n01440764' is the unique synset label associated with
these images.
The training data set consists of 1000 sub-directories (i.e. labels)
each containing 1200 JPEG images for a total of 1.2M JPEG images.
The evaluation data set consists of 1000 sub-directories (i.e. labels)
each containing 50 JPEG images for a total of 50K JPEG images.
This TensorFlow script converts the training and evaluation data into
a sharded data set consisting of 1024 and 128 TFRecord files, respectively.
train_directory/train-00000-of-01024
train_directory/train-00001-of-01024
...
train_directory/train-00127-of-01024
and
validation_directory/validation-00000-of-00128
validation_directory/validation-00001-of-00128
...
validation_directory/validation-00127-of-00128
Each validation TFRecord file contains ~390 records. Each training TFREcord
file contains ~1250 records. Each record within the TFRecord file is a
serialized Example proto. The Example proto contains the following fields:
image/encoded: string containing JPEG encoded image in RGB colorspace
image/height: integer, image height in pixels
image/width: integer, image width in pixels
image/colorspace: string, specifying the colorspace, always 'RGB'
image/channels: integer, specifying the number of channels, always 3
image/format: string, specifying the format, always'JPEG'
image/filename: string containing the basename of the image file
e.g. 'n01440764_10026.JPEG' or 'ILSVRC2012_val_00000293.JPEG'
image/class/label: integer specifying the index in a classification layer.
The label ranges from [1, 1000] where 0 is not used.
image/class/synset: string specifying the unique ID of the label,
e.g. 'n01440764'
image/class/text: string specifying the human-readable version of the label
e.g. 'red fox, Vulpes vulpes'
image/object/bbox/xmin: list of integers specifying the 0+ human annotated
bounding boxes
image/object/bbox/xmax: list of integers specifying the 0+ human annotated
bounding boxes
image/object/bbox/ymin: list of integers specifying the 0+ human annotated
bounding boxes
image/object/bbox/ymax: list of integers specifying the 0+ human annotated
bounding boxes
image/object/bbox/label: integer specifying the index in a classification
layer. The label ranges from [1, 1000] where 0 is not used. Note this is
always identical to the image label.
Note that the length of xmin is identical to the length of xmax, ymin and ymax
for each example.
Running this script using 16 threads may take around ~2.5 hours on a HP Z420.
"""
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
from datetime import datetime
import os
import random
import sys
import threading
import google3
import numpy as np
import tensorflow as tf
tf.app.flags.DEFINE_string('train_directory', '/tmp/',
'Training data directory')
tf.app.flags.DEFINE_string('validation_directory', '/tmp/',
'Validation data directory')
tf.app.flags.DEFINE_string('output_directory', '/tmp/',
'Output data directory')
tf.app.flags.DEFINE_integer('train_shards', 1024,
'Number of shards in training TFRecord files.')
tf.app.flags.DEFINE_integer('validation_shards', 128,
'Number of shards in validation TFRecord files.')
tf.app.flags.DEFINE_integer('num_threads', 8,
'Number of threads to preprocess the images.')
# The labels file contains a list of valid labels are held in this file.
# Assumes that the file contains entries as such:
# n01440764
# n01443537
# n01484850
# where each line corresponds to a label expressed as a synset. We map
# each synset contained in the file to an integer (based on the alphabetical
# ordering). See below for details.
tf.app.flags.DEFINE_string('labels_file',
'imagenet_lsvrc_2015_synsets.txt',
'Labels file')
# This file containing mapping from synset to human-readable label.
# Assumes each line of the file looks like:
#
# n02119247 black fox
# n02119359 silver fox
# n02119477 red fox, Vulpes fulva
#
# where each line corresponds to a unique mapping. Note that each line is
# formatted as <synset>\t<human readable label>.
tf.app.flags.DEFINE_string('imagenet_metadata_file',
'imagenet_metadata.txt',
'ImageNet metadata file')
# This file is the output of process_bounding_box.py
# Assumes each line of the file looks like:
#
# n00007846_64193.JPEG,0.0060,0.2620,0.7545,0.9940
#
# where each line corresponds to one bounding box annotation associated
# with an image. Each line can be parsed as:
#
# <JPEG file name>, <xmin>, <ymin>, <xmax>, <ymax>
#
# Note that there might exist mulitple bounding box annotations associated
# with an image file.
tf.app.flags.DEFINE_string('bounding_box_file',
'./imagenet_2012_bounding_boxes.csv',
'Bounding box file')
FLAGS = tf.app.flags.FLAGS
def _int64_feature(value):
"""Wrapper for inserting int64 features into Example proto."""
if not isinstance(value, list):
value = [value]
return tf.train.Feature(int64_list=tf.train.Int64List(value=value))
def _float_feature(value):
"""Wrapper for inserting float features into Example proto."""
if not isinstance(value, list):
value = [value]
return tf.train.Feature(float_list=tf.train.FloatList(value=value))
def _bytes_feature(value):
"""Wrapper for inserting bytes features into Example proto."""
return tf.train.Feature(bytes_list=tf.train.BytesList(value=[value]))
def _convert_to_example(filename, image_buffer, label, synset, human, bbox,
height, width):
"""Build an Example proto for an example.
Args:
filename: string, path to an image file, e.g., '/path/to/example.JPG'
image_buffer: string, JPEG encoding of RGB image
label: integer, identifier for the ground truth for the network
synset: string, unique WordNet ID specifying the label, e.g., 'n02323233'
human: string, human-readable label, e.g., 'red fox, Vulpes vulpes'
bbox: list of bounding boxes; each box is a list of integers
specifying [xmin, ymin, xmax, ymax]. All boxes are assumed to belong to
the same label as the image label.
height: integer, image height in pixels
width: integer, image width in pixels
Returns:
Example proto
"""
xmin = []
ymin = []
xmax = []
ymax = []
for b in bbox:
assert len(b) == 4
# pylint: disable=expression-not-assigned
[l.append(point) for l, point in zip([xmin, ymin, xmax, ymax], b)]
# pylint: enable=expression-not-assigned
colorspace = 'RGB'
channels = 3
image_format = 'JPEG'
example = tf.train.Example(features=tf.train.Features(feature={
'image/height': _int64_feature(height),
'image/width': _int64_feature(width),
'image/colorspace': _bytes_feature(colorspace),
'image/channels': _int64_feature(channels),
'image/class/label': _int64_feature(label),
'image/class/synset': _bytes_feature(synset),
'image/class/text': _bytes_feature(human),
'image/object/bbox/xmin': _float_feature(xmin),
'image/object/bbox/xmax': _float_feature(xmax),
'image/object/bbox/ymin': _float_feature(ymin),
'image/object/bbox/ymax': _float_feature(ymax),
'image/object/bbox/label': _int64_feature([label] * len(xmin)),
'image/format': _bytes_feature(image_format),
'image/filename': _bytes_feature(os.path.basename(filename)),
'image/encoded': _bytes_feature(image_buffer)}))
return example
class ImageCoder(object):
"""Helper class that provides TensorFlow image coding utilities."""
def __init__(self):
# Create a single Session to run all image coding calls.
self._sess = tf.Session()
# Initializes function that converts PNG to JPEG data.
self._png_data = tf.placeholder(dtype=tf.string)
image = tf.image.decode_png(self._png_data, channels=3)
self._png_to_jpeg = tf.image.encode_jpeg(image, format='rgb', quality=100)
# Initializes function that converts CMYK JPEG data to RGB JPEG data.
self._cmyk_data = tf.placeholder(dtype=tf.string)
image = tf.image.decode_jpeg(self._cmyk_data, channels=0)
self._cmyk_to_rgb = tf.image.encode_jpeg(image, format='rgb', quality=100)
# Initializes function that decodes RGB JPEG data.
self._decode_jpeg_data = tf.placeholder(dtype=tf.string)
self._decode_jpeg = tf.image.decode_jpeg(self._decode_jpeg_data, channels=3)
def png_to_jpeg(self, image_data):
return self._sess.run(self._png_to_jpeg,
feed_dict={self._png_data: image_data})
def cmyk_to_rgb(self, image_data):
return self._sess.run(self._cmyk_to_rgb,
feed_dict={self._cmyk_data: image_data})
def decode_jpeg(self, image_data):
image = self._sess.run(self._decode_jpeg,
feed_dict={self._decode_jpeg_data: image_data})
assert len(image.shape) == 3
assert image.shape[2] == 3
return image
def _is_png(filename):
"""Determine if a file contains a PNG format image.
Args:
filename: string, path of the image file.
Returns:
boolean indicating if the image is a PNG.
"""
# File list from:
# https://groups.google.com/forum/embed/?place=forum/torch7#!topic/torch7/fOSTXHIESSU
return 'n02105855_2933.JPEG' in filename
def _is_cmyk(filename):
"""Determine if file contains a CMYK JPEG format image.
Args:
filename: string, path of the image file.
Returns:
boolean indicating if the image is a JPEG encoded with CMYK color space.
"""
# File list from:
# https://github.com/cytsai/ilsvrc-cmyk-image-list
blacklist = ['n01739381_1309.JPEG', 'n02077923_14822.JPEG',
'n02447366_23489.JPEG', 'n02492035_15739.JPEG',
'n02747177_10752.JPEG', 'n03018349_4028.JPEG',
'n03062245_4620.JPEG', 'n03347037_9675.JPEG',
'n03467068_12171.JPEG', 'n03529860_11437.JPEG',
'n03544143_17228.JPEG', 'n03633091_5218.JPEG',
'n03710637_5125.JPEG', 'n03961711_5286.JPEG',
'n04033995_2932.JPEG', 'n04258138_17003.JPEG',
'n04264628_27969.JPEG', 'n04336792_7448.JPEG',
'n04371774_5854.JPEG', 'n04596742_4225.JPEG',
'n07583066_647.JPEG', 'n13037406_4650.JPEG']
return filename.split('/')[-1] in blacklist
def _process_image(filename, coder):
"""Process a single image file.
Args:
filename: string, path to an image file e.g., '/path/to/example.JPG'.
coder: instance of ImageCoder to provide TensorFlow image coding utils.
Returns:
image_buffer: string, JPEG encoding of RGB image.
height: integer, image height in pixels.
width: integer, image width in pixels.
"""
# Read the image file.
image_data = tf.gfile.FastGFile(filename, 'r').read()
# Clean the dirty data.
if _is_png(filename):
# 1 image is a PNG.
print('Converting PNG to JPEG for %s' % filename)
image_data = coder.png_to_jpeg(image_data)
elif _is_cmyk(filename):
# 22 JPEG images are in CMYK colorspace.
print('Converting CMYK to RGB for %s' % filename)
image_data = coder.cmyk_to_rgb(image_data)
# Decode the RGB JPEG.
image = coder.decode_jpeg(image_data)
# Check that image converted to RGB
assert len(image.shape) == 3
height = image.shape[0]
width = image.shape[1]
assert image.shape[2] == 3
return image_data, height, width
def _process_image_files_batch(coder, thread_index, ranges, name, filenames,
synsets, labels, humans, bboxes, num_shards):
"""Processes and saves list of images as TFRecord in 1 thread.
Args:
coder: instance of ImageCoder to provide TensorFlow image coding utils.
thread_index: integer, unique batch to run index is within [0, len(ranges)).
ranges: list of pairs of integers specifying ranges of each batches to
analyze in parallel.
name: string, unique identifier specifying the data set
filenames: list of strings; each string is a path to an image file
synsets: list of strings; each string is a unique WordNet ID
labels: list of integer; each integer identifies the ground truth
humans: list of strings; each string is a human-readable label
bboxes: list of bounding boxes for each image. Note that each entry in this
list might contain from 0+ entries corresponding to the number of bounding
box annotations for the image.
num_shards: integer number of shards for this data set.
"""
# Each thread produces N shards where N = int(num_shards / num_threads).
# For instance, if num_shards = 128, and the num_threads = 2, then the first
# thread would produce shards [0, 64).
num_threads = len(ranges)
assert not num_shards % num_threads
num_shards_per_batch = int(num_shards / num_threads)
shard_ranges = np.linspace(ranges[thread_index][0],
ranges[thread_index][1],
num_shards_per_batch + 1).astype(int)
num_files_in_thread = ranges[thread_index][1] - ranges[thread_index][0]
counter = 0
for s in xrange(num_shards_per_batch):
# Generate a sharded version of the file name, e.g. 'train-00002-of-00010'
shard = thread_index * num_shards_per_batch + s
output_filename = '%s-%.5d-of-%.5d' % (name, shard, num_shards)
output_file = os.path.join(FLAGS.output_directory, output_filename)
writer = tf.python_io.TFRecordWriter(output_file)
shard_counter = 0
files_in_shard = np.arange(shard_ranges[s], shard_ranges[s + 1], dtype=int)
for i in files_in_shard:
filename = filenames[i]
label = labels[i]
synset = synsets[i]
human = humans[i]
bbox = bboxes[i]
image_buffer, height, width = _process_image(filename, coder)
example = _convert_to_example(filename, image_buffer, label,
synset, human, bbox,
height, width)
writer.write(example.SerializeToString())
shard_counter += 1
counter += 1
if not counter % 1000:
print('%s [thread %d]: Processed %d of %d images in thread batch.' %
(datetime.now(), thread_index, counter, num_files_in_thread))
sys.stdout.flush()
writer.close()
print('%s [thread %d]: Wrote %d images to %s' %
(datetime.now(), thread_index, shard_counter, output_file))
sys.stdout.flush()
shard_counter = 0
print('%s [thread %d]: Wrote %d images to %d shards.' %
(datetime.now(), thread_index, counter, num_files_in_thread))
sys.stdout.flush()
def _process_image_files(name, filenames, synsets, labels, humans,
bboxes, num_shards):
"""Process and save list of images as TFRecord of Example protos.
Args:
name: string, unique identifier specifying the data set
filenames: list of strings; each string is a path to an image file
synsets: list of strings; each string is a unique WordNet ID
labels: list of integer; each integer identifies the ground truth
humans: list of strings; each string is a human-readable label
bboxes: list of bounding boxes for each image. Note that each entry in this
list might contain from 0+ entries corresponding to the number of bounding
box annotations for the image.
num_shards: integer number of shards for this data set.
"""
assert len(filenames) == len(synsets)
assert len(filenames) == len(labels)
assert len(filenames) == len(humans)
assert len(filenames) == len(bboxes)
# Break all images into batches with a [ranges[i][0], ranges[i][1]].
spacing = np.linspace(0, len(filenames), FLAGS.num_threads + 1).astype(np.int)
ranges = []
threads = []
for i in xrange(len(spacing) - 1):
ranges.append([spacing[i], spacing[i+1]])
# Launch a thread for each batch.
print('Launching %d threads for spacings: %s' % (FLAGS.num_threads, ranges))
sys.stdout.flush()
# Create a mechanism for monitoring when all threads are finished.
coord = tf.train.Coordinator()
# Create a generic TensorFlow-based utility for converting all image codings.
coder = ImageCoder()
threads = []
for thread_index in xrange(len(ranges)):
args = (coder, thread_index, ranges, name, filenames,
synsets, labels, humans, bboxes, num_shards)
t = threading.Thread(target=_process_image_files_batch, args=args)
t.start()
threads.append(t)
# Wait for all the threads to terminate.
coord.join(threads)
print('%s: Finished writing all %d images in data set.' %
(datetime.now(), len(filenames)))
sys.stdout.flush()
def _find_image_files(data_dir, labels_file):
"""Build a list of all images files and labels in the data set.
Args:
data_dir: string, path to the root directory of images.
Assumes that the ImageNet data set resides in JPEG files located in
the following directory structure.
data_dir/n01440764/ILSVRC2012_val_00000293.JPEG
data_dir/n01440764/ILSVRC2012_val_00000543.JPEG
where 'n01440764' is the unique synset label associated with these images.
labels_file: string, path to the labels file.
The list of valid labels are held in this file. Assumes that the file
contains entries as such:
n01440764
n01443537
n01484850
where each line corresponds to a label expressed as a synset. We map
each synset contained in the file to an integer (based on the alphabetical
ordering) starting with the integer 1 corresponding to the synset
contained in the first line.
The reason we start the integer labels at 1 is to reserve label 0 as an
unused background class.
Returns:
filenames: list of strings; each string is a path to an image file.
synsets: list of strings; each string is a unique WordNet ID.
labels: list of integer; each integer identifies the ground truth.
"""
print('Determining list of input files and labels from %s.' % data_dir)
challenge_synsets = [l.strip() for l in
tf.gfile.FastGFile(labels_file, 'r').readlines()]
labels = []
filenames = []
synsets = []
# Leave label index 0 empty as a background class.
label_index = 1
# Construct the list of JPEG files and labels.
for synset in challenge_synsets:
jpeg_file_path = '%s/%s/*.JPEG' % (data_dir, synset)
matching_files = tf.gfile.Glob(jpeg_file_path)
labels.extend([label_index] * len(matching_files))
synsets.extend([synset] * len(matching_files))
filenames.extend(matching_files)
if not label_index % 100:
print('Finished finding files in %d of %d classes.' % (
label_index, len(challenge_synsets)))
label_index += 1
# Shuffle the ordering of all image files in order to guarantee
# random ordering of the images with respect to label in the
# saved TFRecord files. Make the randomization repeatable.
shuffled_index = range(len(filenames))
random.seed(12345)
random.shuffle(shuffled_index)
filenames = [filenames[i] for i in shuffled_index]
synsets = [synsets[i] for i in shuffled_index]
labels = [labels[i] for i in shuffled_index]
print('Found %d JPEG files across %d labels inside %s.' %
(len(filenames), len(challenge_synsets), data_dir))
return filenames, synsets, labels
def _find_human_readable_labels(synsets, synset_to_human):
"""Build a list of human-readable labels.
Args:
synsets: list of strings; each string is a unique WordNet ID.
synset_to_human: dict of synset to human labels, e.g.,
'n02119022' --> 'red fox, Vulpes vulpes'
Returns:
List of human-readable strings corresponding to each synset.
"""
humans = []
for s in synsets:
assert s in synset_to_human, ('Failed to find: %s' % s)
humans.append(synset_to_human[s])
return humans
def _find_image_bounding_boxes(filenames, image_to_bboxes):
"""Find the bounding boxes for a given image file.
Args:
filenames: list of strings; each string is a path to an image file.
image_to_bboxes: dictionary mapping image file names to a list of
bounding boxes. This list contains 0+ bounding boxes.
Returns:
List of bounding boxes for each image. Note that each entry in this
list might contain from 0+ entries corresponding to the number of bounding
box annotations for the image.
"""
num_image_bbox = 0
bboxes = []
for f in filenames:
basename = os.path.basename(f)
if basename in image_to_bboxes:
bboxes.append(image_to_bboxes[basename])
num_image_bbox += 1
else:
bboxes.append([])
print('Found %d images with bboxes out of %d images' % (
num_image_bbox, len(filenames)))
return bboxes
def _process_dataset(name, directory, num_shards, synset_to_human,
image_to_bboxes):
"""Process a complete data set and save it as a TFRecord.
Args:
name: string, unique identifier specifying the data set.
directory: string, root path to the data set.
num_shards: integer number of shards for this data set.
synset_to_human: dict of synset to human labels, e.g.,
'n02119022' --> 'red fox, Vulpes vulpes'
image_to_bboxes: dictionary mapping image file names to a list of
bounding boxes. This list contains 0+ bounding boxes.
"""
filenames, synsets, labels = _find_image_files(directory, FLAGS.labels_file)
humans = _find_human_readable_labels(synsets, synset_to_human)
bboxes = _find_image_bounding_boxes(filenames, image_to_bboxes)
_process_image_files(name, filenames, synsets, labels,
humans, bboxes, num_shards)
def _build_synset_lookup(imagenet_metadata_file):
"""Build lookup for synset to human-readable label.
Args:
imagenet_metadata_file: string, path to file containing mapping from
synset to human-readable label.
Assumes each line of the file looks like:
n02119247 black fox
n02119359 silver fox
n02119477 red fox, Vulpes fulva
where each line corresponds to a unique mapping. Note that each line is
formatted as <synset>\t<human readable label>.
Returns:
Dictionary of synset to human labels, such as:
'n02119022' --> 'red fox, Vulpes vulpes'
"""
lines = tf.gfile.FastGFile(imagenet_metadata_file, 'r').readlines()
synset_to_human = {}
for l in lines:
if l:
parts = l.strip().split('\t')
assert len(parts) == 2
synset = parts[0]
human = parts[1]
synset_to_human[synset] = human
return synset_to_human
def _build_bounding_box_lookup(bounding_box_file):
"""Build a lookup from image file to bounding boxes.
Args:
bounding_box_file: string, path to file with bounding boxes annotations.
Assumes each line of the file looks like:
n00007846_64193.JPEG,0.0060,0.2620,0.7545,0.9940
where each line corresponds to one bounding box annotation associated
with an image. Each line can be parsed as:
<JPEG file name>, <xmin>, <ymin>, <xmax>, <ymax>
Note that there might exist mulitple bounding box annotations associated
with an image file. This file is the output of process_bounding_boxes.py.
Returns:
Dictionary mapping image file names to a list of bounding boxes. This list
contains 0+ bounding boxes.
"""
lines = tf.gfile.FastGFile(bounding_box_file, 'r').readlines()
images_to_bboxes = {}
num_bbox = 0
num_image = 0
for l in lines:
if l:
parts = l.split(',')
assert len(parts) == 5, ('Failed to parse: %s' % l)
filename = parts[0]
xmin = float(parts[1])
ymin = float(parts[2])
xmax = float(parts[3])
ymax = float(parts[4])
box = [xmin, ymin, xmax, ymax]
if filename not in images_to_bboxes:
images_to_bboxes[filename] = []
num_image += 1
images_to_bboxes[filename].append(box)
num_bbox += 1
print('Successfully read %d bounding boxes '
'across %d images.' % (num_bbox, num_image))
return images_to_bboxes
def main(unused_argv):
assert not FLAGS.train_shards % FLAGS.num_threads, (
'Please make the FLAGS.num_threads commensurate with FLAGS.train_shards')
assert not FLAGS.validation_shards % FLAGS.num_threads, (
'Please make the FLAGS.num_threads commensurate with '
'FLAGS.validation_shards')
print('Saving results to %s' % FLAGS.output_directory)
# Build a map from synset to human-readable label.
synset_to_human = _build_synset_lookup(FLAGS.imagenet_metadata_file)
image_to_bboxes = _build_bounding_box_lookup(FLAGS.bounding_box_file)
# Run it!
_process_dataset('validation', FLAGS.validation_directory,
FLAGS.validation_shards, synset_to_human, image_to_bboxes)
_process_dataset('train', FLAGS.train_directory, FLAGS.train_shards,
synset_to_human, image_to_bboxes)
if __name__ == '__main__':
tf.app.run()
slim/datasets/download_and_convert_imagenet.sh 0 → 100755
View file @ 70b894bd
#!/bin/bash
# Copyright 2016 Google Inc. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# ==============================================================================
# Script to download and preprocess ImageNet Challenge 2012
# training and validation data set.
#
# The final output of this script are sharded TFRecord files containing
# serialized Example protocol buffers. See build_imagenet_data.py for
# details of how the Example protocol buffers contain the ImageNet data.
#
# The final output of this script appears as such:
#
# data_dir/train-00000-of-01024
# data_dir/train-00001-of-01024
# ...
# data_dir/train-00127-of-01024
#
# and
#
# data_dir/validation-00000-of-00128
# data_dir/validation-00001-of-00128
# ...
# data_dir/validation-00127-of-00128
#
# Note that this script may take several hours to run to completion. The
# conversion of the ImageNet data to TFRecords alone takes 2-3 hours depending
# on the speed of your machine. Please be patient.
#
# **IMPORTANT**
# To download the raw images, the user must create an account with image-net.org
# and generate a username and access_key. The latter two are required for
# downloading the raw images.
#
# usage:
# ./download_and_preprocess_imagenet.sh [data-dir]
set -e
if [ -z "$1" ]; then
echo "usage download_and_preprocess_imagenet.sh [data dir]"
exit
fi
# Create the output and temporary directories.
DATA_DIR="${1%/}"
SCRATCH_DIR="${DATA_DIR}/raw-data/"
mkdir -p "${DATA_DIR}"
mkdir -p "${SCRATCH_DIR}"
WORK_DIR="$0.runfiles/third_party/tensorflow_models/slim"
# Download the ImageNet data.
LABELS_FILE="${WORK_DIR}/datasets/imagenet_lsvrc_2015_synsets.txt"
DOWNLOAD_SCRIPT="${WORK_DIR}/datasets/download_imagenet.sh"
"${DOWNLOAD_SCRIPT}" "${SCRATCH_DIR}" "${LABELS_FILE}"
# Note the locations of the train and validation data.
TRAIN_DIRECTORY="${SCRATCH_DIR}train/"
VALIDATION_DIRECTORY="${SCRATCH_DIR}validation/"
# Preprocess the validation data by moving the images into the appropriate
# sub-directory based on the label (synset) of the image.
echo "Organizing the validation data into sub-directories."
PREPROCESS_VAL_SCRIPT="${WORK_DIR}/datasets/preprocess_imagenet_validation_data.py"
VAL_LABELS_FILE="${WORK_DIR}/datasets/imagenet_2012_validation_synset_labels.txt"
"${PREPROCESS_VAL_SCRIPT}" "${VALIDATION_DIRECTORY}" "${VAL_LABELS_FILE}"
# Convert the XML files for bounding box annotations into a single CSV.
echo "Extracting bounding box information from XML."
BOUNDING_BOX_SCRIPT="${WORK_DIR}/datasets/process_bounding_boxes.py"
BOUNDING_BOX_FILE="${SCRATCH_DIR}/imagenet_2012_bounding_boxes.csv"
BOUNDING_BOX_DIR="${SCRATCH_DIR}bounding_boxes/"
"${BOUNDING_BOX_SCRIPT}" "${BOUNDING_BOX_DIR}" "${LABELS_FILE}" \
| sort >"${BOUNDING_BOX_FILE}"
echo "Finished downloading and preprocessing the ImageNet data."
# Build the TFRecords version of the ImageNet data.
BUILD_SCRIPT="${WORK_DIR}/build_imagenet_data"
OUTPUT_DIRECTORY="${DATA_DIR}"
IMAGENET_METADATA_FILE="${WORK_DIR}/datasets/imagenet_metadata.txt"
"${BUILD_SCRIPT}" \
--train_directory="${TRAIN_DIRECTORY}" \
--validation_directory="${VALIDATION_DIRECTORY}" \
--output_directory="${OUTPUT_DIRECTORY}" \
--imagenet_metadata_file="${IMAGENET_METADATA_FILE}" \
--labels_file="${LABELS_FILE}" \
--bounding_box_file="${BOUNDING_BOX_FILE}"
slim/datasets/download_imagenet.sh 0 → 100755
View file @ 70b894bd
#!/bin/bash
# Copyright 2016 Google Inc. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# ==============================================================================
# Script to download ImageNet Challenge 2012 training and validation data set.
#
# Downloads and decompresses raw images and bounding boxes.
#
# **IMPORTANT**
# To download the raw images, the user must create an account with image-net.org
# and generate a username and access_key. The latter two are required for
# downloading the raw images.
#
# usage:
# ./download_imagenet.sh [dirname]
set -e
if [ "x$IMAGENET_ACCESS_KEY" == x -o "x$IMAGENET_USERNAME" == x ]; then
cat <<END
In order to download the imagenet data, you have to create an account with
image-net.org. This will get you a username and an access key. You can set the
IMAGENET_USERNAME and IMAGENET_ACCESS_KEY environment variables, or you can
enter the credentials here.
END
read -p "Username: " IMAGENET_USERNAME
read -p "Access key: " IMAGENET_ACCESS_KEY
fi
OUTDIR="${1:-./imagenet-data}"
SYNSETS_FILE="${2:-./synsets.txt}"
echo "Saving downloaded files to $OUTDIR"
mkdir -p "${OUTDIR}"
CURRENT_DIR=$(pwd)
BBOX_DIR="${OUTDIR}bounding_boxes"
mkdir -p "${BBOX_DIR}"
cd "${OUTDIR}"
# Download and process all of the ImageNet bounding boxes.
BASE_URL="http://www.image-net.org/challenges/LSVRC/2012/nonpub"
# See here for details: http://www.image-net.org/download-bboxes
BOUNDING_BOX_ANNOTATIONS="${BASE_URL}/ILSVRC2012_bbox_train_v2.tar.gz"
BBOX_TAR_BALL="${BBOX_DIR}/annotations.tar.gz"
echo "Downloading bounding box annotations."
wget "${BOUNDING_BOX_ANNOTATIONS}" -O "${BBOX_TAR_BALL}"
echo "Uncompressing bounding box annotations ..."
tar xzf "${BBOX_TAR_BALL}" -C "${BBOX_DIR}"
LABELS_ANNOTATED="${BBOX_DIR}/*"
NUM_XML=$(ls -1 ${LABELS_ANNOTATED} | wc -l)
echo "Identified ${NUM_XML} bounding box annotations."
# Download and uncompress all images from the ImageNet 2012 validation dataset.
VALIDATION_TARBALL="ILSVRC2012_img_val.tar"
OUTPUT_PATH="${OUTDIR}validation/"
mkdir -p "${OUTPUT_PATH}"
cd "${OUTDIR}/.."
echo "Downloading ${VALIDATION_TARBALL} to ${OUTPUT_PATH}."
wget -nd -c "${BASE_URL}/${VALIDATION_TARBALL}"
tar xf "${VALIDATION_TARBALL}" -C "${OUTPUT_PATH}"
# Download all images from the ImageNet 2012 train dataset.
TRAIN_TARBALL="ILSVRC2012_img_train.tar"
OUTPUT_PATH="${OUTDIR}train/"
mkdir -p "${OUTPUT_PATH}"
cd "${OUTDIR}/.."
echo "Downloading ${TRAIN_TARBALL} to ${OUTPUT_PATH}."
wget -nd -c "${BASE_URL}/${TRAIN_TARBALL}"
# Un-compress the individual tar-files within the train tar-file.
echo "Uncompressing individual train tar-balls in the training data."
while read SYNSET; do
echo "Processing: ${SYNSET}"
# Create a directory and delete anything there.
mkdir -p "${OUTPUT_PATH}/${SYNSET}"
rm -rf "${OUTPUT_PATH}/${SYNSET}/*"
# Uncompress into the directory.
tar xf "${TRAIN_TARBALL}" "${SYNSET}.tar"
tar xf "${SYNSET}.tar" -C "${OUTPUT_PATH}/${SYNSET}/"
rm -f "${SYNSET}.tar"
echo "Finished processing: ${SYNSET}"
done < "${SYNSETS_FILE}"
slim/datasets/imagenet_2012_validation_synset_labels.txt 0 → 100644
View file @ 70b894bd
This source diff could not be displayed because it is too large. You can view the blob instead.
slim/datasets/imagenet_lsvrc_2015_synsets.txt 0 → 100644
View file @ 70b894bd
n01440764
n01443537
n01484850
n01491361
n01494475
n01496331
n01498041
n01514668
n01514859
n01518878
n01530575
n01531178
n01532829
n01534433
n01537544
n01558993
n01560419
n01580077
n01582220
n01592084
n01601694
n01608432
n01614925
n01616318
n01622779
n01629819
n01630670
n01631663
n01632458
n01632777
n01641577
n01644373
n01644900
n01664065
n01665541
n01667114
n01667778
n01669191
n01675722
n01677366
n01682714
n01685808
n01687978
n01688243
n01689811
n01692333
n01693334
n01694178
n01695060
n01697457
n01698640
n01704323
n01728572
n01728920
n01729322
n01729977
n01734418
n01735189
n01737021
n01739381
n01740131
n01742172
n01744401
n01748264
n01749939
n01751748
n01753488
n01755581
n01756291
n01768244
n01770081
n01770393
n01773157
n01773549
n01773797
n01774384
n01774750
n01775062
n01776313
n01784675
n01795545
n01796340
n01797886
n01798484
n01806143
n01806567
n01807496
n01817953
n01818515
n01819313
n01820546
n01824575
n01828970
n01829413
n01833805
n01843065
n01843383
n01847000
n01855032
n01855672
n01860187
n01871265
n01872401
n01873310
n01877812
n01882714
n01883070
n01910747
n01914609
n01917289
n01924916
n01930112
n01943899
n01944390
n01945685
n01950731
n01955084
n01968897
n01978287
n01978455
n01980166
n01981276
n01983481
n01984695
n01985128
n01986214
n01990800
n02002556
n02002724
n02006656
n02007558
n02009229
n02009912
n02011460
n02012849
n02013706
n02017213
n02018207
n02018795
n02025239
n02027492
n02028035
n02033041
n02037110
n02051845
n02056570
n02058221
n02066245
n02071294
n02074367
n02077923
n02085620
n02085782
n02085936
n02086079
n02086240
n02086646
n02086910
n02087046
n02087394
n02088094
n02088238
n02088364
n02088466
n02088632
n02089078
n02089867
n02089973
n02090379
n02090622
n02090721
n02091032
n02091134
n02091244
n02091467
n02091635
n02091831
n02092002
n02092339
n02093256
n02093428
n02093647
n02093754
n02093859
n02093991
n02094114
n02094258
n02094433
n02095314
n02095570
n02095889
n02096051
n02096177
n02096294
n02096437
n02096585
n02097047
n02097130
n02097209
n02097298
n02097474
n02097658
n02098105
n02098286
n02098413
n02099267
n02099429
n02099601
n02099712
n02099849
n02100236
n02100583
n02100735
n02100877
n02101006
n02101388
n02101556
n02102040
n02102177
n02102318
n02102480
n02102973
n02104029
n02104365
n02105056
n02105162
n02105251
n02105412
n02105505
n02105641
n02105855
n02106030
n02106166
n02106382
n02106550
n02106662
n02107142
n02107312
n02107574
n02107683
n02107908
n02108000
n02108089
n02108422
n02108551
n02108915
n02109047
n02109525
n02109961
n02110063
n02110185
n02110341
n02110627
n02110806
n02110958
n02111129
n02111277
n02111500
n02111889
n02112018
n02112137
n02112350
n02112706
n02113023
n02113186
n02113624
n02113712
n02113799
n02113978
n02114367
n02114548
n02114712
n02114855
n02115641
n02115913
n02116738
n02117135
n02119022
n02119789
n02120079
n02120505
n02123045
n02123159
n02123394
n02123597
n02124075
n02125311
n02127052
n02128385
n02128757
n02128925
n02129165
n02129604
n02130308
n02132136
n02133161
n02134084
n02134418
n02137549
n02138441
n02165105
n02165456
n02167151
n02168699
n02169497
n02172182
n02174001
n02177972
n02190166
n02206856
n02219486
n02226429
n02229544
n02231487
n02233338
n02236044
n02256656
n02259212
n02264363
n02268443
n02268853
n02276258
n02277742
n02279972
n02280649
n02281406
n02281787
n02317335
n02319095
n02321529
n02325366
n02326432
n02328150
n02342885
n02346627
n02356798
n02361337
n02363005
n02364673
n02389026
n02391049
n02395406
n02396427
n02397096
n02398521
n02403003
n02408429
n02410509
n02412080
n02415577
n02417914
n02422106
n02422699
n02423022
n02437312
n02437616
n02441942
n02442845
n02443114
n02443484
n02444819
n02445715
n02447366
n02454379
n02457408
n02480495
n02480855
n02481823
n02483362
n02483708
n02484975
n02486261
n02486410
n02487347
n02488291
n02488702
n02489166
n02490219
n02492035
n02492660
n02493509
n02493793
n02494079
n02497673
n02500267
n02504013
n02504458
n02509815
n02510455
n02514041
n02526121
n02536864
n02606052
n02607072
n02640242
n02641379
n02643566
n02655020
n02666196
n02667093
n02669723
n02672831
n02676566
n02687172
n02690373
n02692877
n02699494
n02701002
n02704792
n02708093
n02727426
n02730930
n02747177
n02749479
n02769748
n02776631
n02777292
n02782093
n02783161
n02786058
n02787622
n02788148
n02790996
n02791124
n02791270
n02793495
n02794156
n02795169
n02797295
n02799071
n02802426
n02804414
n02804610
n02807133
n02808304
n02808440
n02814533
n02814860
n02815834
n02817516
n02823428
n02823750
n02825657
n02834397
n02835271
n02837789
n02840245
n02841315
n02843684
n02859443
n02860847
n02865351
n02869837
n02870880
n02871525
n02877765
n02879718
n02883205
n02892201
n02892767
n02894605
n02895154
n02906734
n02909870
n02910353
n02916936
n02917067
n02927161
n02930766
n02939185
n02948072
n02950826
n02951358
n02951585
n02963159
n02965783
n02966193
n02966687
n02971356
n02974003
n02977058
n02978881
n02979186
n02980441
n02981792
n02988304
n02992211
n02992529
n02999410
n03000134
n03000247
n03000684
n03014705
n03016953
n03017168
n03018349
n03026506
n03028079
n03032252
n03041632
n03042490
n03045698
n03047690
n03062245
n03063599
n03063689
n03065424
n03075370
n03085013
n03089624
n03095699
n03100240
n03109150
n03110669
n03124043
n03124170
n03125729
n03126707
n03127747
n03127925
n03131574
n03133878
n03134739
n03141823
n03146219
n03160309
n03179701
n03180011
n03187595
n03188531
n03196217
n03197337
n03201208
n03207743
n03207941
n03208938
n03216828
n03218198
n03220513
n03223299
n03240683
n03249569
n03250847
n03255030
n03259280
n03271574
n03272010
n03272562
n03290653
n03291819
n03297495
n03314780
n03325584
n03337140
n03344393
n03345487
n03347037
n03355925
n03372029
n03376595
n03379051
n03384352
n03388043
n03388183
n03388549
n03393912
n03394916
n03400231
n03404251
n03417042
n03424325
n03425413
n03443371
n03444034
n03445777
n03445924
n03447447
n03447721
n03450230
n03452741
n03457902
n03459775
n03461385
n03467068
n03476684
n03476991
n03478589
n03481172
n03482405
n03483316
n03485407
n03485794
n03492542
n03494278
n03495258
n03496892
n03498962
n03527444
n03529860
n03530642
n03532672
n03534580
n03535780
n03538406
n03544143
n03584254
n03584829
n03590841
n03594734
n03594945
n03595614
n03598930
n03599486
n03602883
n03617480
n03623198
n03627232
n03630383
n03633091
n03637318
n03642806
n03649909
n03657121
n03658185
n03661043
n03662601
n03666591
n03670208
n03673027
n03676483
n03680355
n03690938
n03691459
n03692522
n03697007
n03706229
n03709823
n03710193
n03710637
n03710721
n03717622
n03720891
n03721384
n03724870
n03729826
n03733131
n03733281
n03733805
n03742115
n03743016
n03759954
n03761084
n03763968
n03764736
n03769881
n03770439
n03770679
n03773504
n03775071
n03775546
n03776460
n03777568
n03777754
n03781244
n03782006
n03785016
n03786901
n03787032
n03788195
n03788365
n03791053
n03792782
n03792972
n03793489
n03794056
n03796401
n03803284
n03804744
n03814639
n03814906
n03825788
n03832673
n03837869
n03838899
n03840681
n03841143
n03843555
n03854065
n03857828
n03866082
n03868242
n03868863
n03871628
n03873416
n03874293
n03874599
n03876231
n03877472
n03877845
n03884397
n03887697
n03888257
n03888605
n03891251
n03891332
n03895866
n03899768
n03902125
n03903868
n03908618
n03908714
n03916031
n03920288
n03924679
n03929660
n03929855
n03930313
n03930630
n03933933
n03935335
n03937543
n03938244
n03942813
n03944341
n03947888
n03950228
n03954731
n03956157
n03958227
n03961711
n03967562
n03970156
n03976467
n03976657
n03977966
n03980874
n03982430
n03983396
n03991062
n03992509
n03995372
n03998194
n04004767
n04005630
n04008634
n04009552
n04019541
n04023962
n04026417
n04033901
n04033995
n04037443
n04039381
n04040759
n04041544
n04044716
n04049303
n04065272
n04067472
n04069434
n04070727
n04074963
n04081281
n04086273
n04090263
n04099969
n04111531
n04116512
n04118538
n04118776
n04120489
n04125021
n04127249
n04131690
n04133789
n04136333
n04141076
n04141327
n04141975
n04146614
n04147183
n04149813
n04152593
n04153751
n04154565
n04162706
n04179913
n04192698
n04200800
n04201297
n04204238
n04204347
n04208210
n04209133
n04209239
n04228054
n04229816
n04235860
n04238763
n04239074
n04243546
n04251144
n04252077
n04252225
n04254120
n04254680
n04254777
n04258138
n04259630
n04263257
n04264628
n04265275
n04266014
n04270147
n04273569
n04275548
n04277352
n04285008
n04286575
n04296562
n04310018
n04311004
n04311174
n04317175
n04325704
n04326547
n04328186
n04330267
n04332243
n04335435
n04336792
n04344873
n04346328
n04347754
n04350905
n04355338
n04355933
n04356056
n04357314
n04366367
n04367480
n04370456
n04371430
n04371774
n04372370
n04376876
n04380533
n04389033
n04392985
n04398044
n04399382
n04404412
n04409515
n04417672
n04418357
n04423845
n04428191
n04429376
n04435653
n04442312
n04443257
n04447861
n04456115
n04458633
n04461696
n04462240
n04465501
n04467665
n04476259
n04479046
n04482393
n04483307
n04485082
n04486054
n04487081
n04487394
n04493381
n04501370
n04505470
n04507155
n04509417
n04515003
n04517823
n04522168
n04523525
n04525038
n04525305
n04532106
n04532670
n04536866
n04540053
n04542943
n04548280
n04548362
n04550184
n04552348
n04553703
n04554684
n04557648
n04560804
n04562935
n04579145
n04579432
n04584207
n04589890
n04590129
n04591157
n04591713
n04592741
n04596742
n04597913
n04599235
n04604644
n04606251
n04612504
n04613696
n06359193
n06596364
n06785654
n06794110
n06874185
n07248320
n07565083
n07579787
n07583066
n07584110
n07590611
n07613480
n07614500
n07615774
n07684084
n07693725
n07695742
n07697313
n07697537
n07711569
n07714571
n07714990
n07715103
n07716358
n07716906
n07717410
n07717556
n07718472
n07718747
n07720875
n07730033
n07734744
n07742313
n07745940
n07747607
n07749582
n07753113
n07753275
n07753592
n07754684
n07760859
n07768694
n07802026
n07831146
n07836838
n07860988
n07871810
n07873807
n07875152
n07880968
n07892512
n07920052
n07930864
n07932039
n09193705
n09229709
n09246464
n09256479
n09288635
n09332890
n09399592
n09421951
n09428293
n09468604
n09472597
n09835506
n10148035
n10565667
n11879895
n11939491
n12057211
n12144580
n12267677
n12620546
n12768682
n12985857
n12998815
n13037406
n13040303
n13044778
n13052670
n13054560
n13133613
n15075141
slim/datasets/imagenet_metadata.txt 0 → 100644
View file @ 70b894bd
This source diff could not be displayed because it is too large. You can view the blob instead.
slim/datasets/preprocess_imagenet_validation_data.py 0 → 100755
View file @ 70b894bd
#!/usr/bin/python
# Copyright 2016 Google Inc. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# ==============================================================================
"""Process the ImageNet Challenge bounding boxes for TensorFlow model training.
Associate the ImageNet 2012 Challenge validation data set with labels.
The raw ImageNet validation data set is expected to reside in JPEG files
located in the following directory structure.
data_dir/ILSVRC2012_val_00000001.JPEG
data_dir/ILSVRC2012_val_00000002.JPEG
...
data_dir/ILSVRC2012_val_00050000.JPEG
This script moves the files into a directory structure like such:
data_dir/n01440764/ILSVRC2012_val_00000293.JPEG
data_dir/n01440764/ILSVRC2012_val_00000543.JPEG
...
where 'n01440764' is the unique synset label associated with
these images.
This directory reorganization requires a mapping from validation image
number (i.e. suffix of the original file) to the associated label. This
is provided in the ImageNet development kit via a Matlab file.
In order to make life easier and divorce ourselves from Matlab, we instead
supply a custom text file that provides this mapping for us.
Sample usage:
./preprocess_imagenet_validation_data.py ILSVRC2012_img_val \
imagenet_2012_validation_synset_labels.txt
"""
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
import os
import os.path
import sys
if __name__ == '__main__':
if len(sys.argv) < 3:
print('Invalid usage\n'
'usage: preprocess_imagenet_validation_data.py '
'<validation data dir> <validation labels file>')
sys.exit(-1)
data_dir = sys.argv[1]
validation_labels_file = sys.argv[2]
# Read in the 50000 synsets associated with the validation data set.
labels = [l.strip() for l in open(validation_labels_file).readlines()]
unique_labels = set(labels)
# Make all sub-directories in the validation data dir.
for label in unique_labels:
labeled_data_dir = os.path.join(data_dir, label)
os.makedirs(labeled_data_dir)
# Move all of the image to the appropriate sub-directory.
for i in xrange(len(labels)):
basename = 'ILSVRC2012_val_000%.5d.JPEG' % (i + 1)
original_filename = os.path.join(data_dir, basename)
if not os.path.exists(original_filename):
print('Failed to find: ' % original_filename)
sys.exit(-1)
new_filename = os.path.join(data_dir, labels[i], basename)
os.rename(original_filename, new_filename)
slim/datasets/process_bounding_boxes.py 0 → 100755
View file @ 70b894bd
#!/usr/bin/python
# Copyright 2016 Google Inc. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# ==============================================================================
"""Process the ImageNet Challenge bounding boxes for TensorFlow model training.
This script is called as
process_bounding_boxes.py <dir> [synsets-file]
Where <dir> is a directory containing the downloaded and unpacked bounding box
data. If [synsets-file] is supplied, then only the bounding boxes whose
synstes are contained within this file are returned. Note that the
[synsets-file] file contains synset ids, one per line.
The script dumps out a CSV text file in which each line contains an entry.
n00007846_64193.JPEG,0.0060,0.2620,0.7545,0.9940
The entry can be read as:
<JPEG file name>, <xmin>, <ymin>, <xmax>, <ymax>
The bounding box for <JPEG file name> contains two points (xmin, ymin) and
(xmax, ymax) specifying the lower-left corner and upper-right corner of a
bounding box in *relative* coordinates.
The user supplies a directory where the XML files reside. The directory
structure in the directory <dir> is assumed to look like this:
<dir>/nXXXXXXXX/nXXXXXXXX_YYYY.xml
Each XML file contains a bounding box annotation. The script:
(1) Parses the XML file and extracts the filename, label and bounding box info.
(2) The bounding box is specified in the XML files as integer (xmin, ymin) and
(xmax, ymax) *relative* to image size displayed to the human annotator. The
size of the image displayed to the human annotator is stored in the XML file
as integer (height, width).
Note that the displayed size will differ from the actual size of the image
downloaded from image-net.org. To make the bounding box annotation useable,
we convert bounding box to floating point numbers relative to displayed
height and width of the image.
Note that each XML file might contain N bounding box annotations.
Note that the points are all clamped at a range of [0.0, 1.0] because some
human annotations extend outside the range of the supplied image.
See details here: http://image-net.org/download-bboxes
(3) By default, the script outputs all valid bounding boxes. If a
[synsets-file] is supplied, only the subset of bounding boxes associated
with those synsets are outputted. Importantly, one can supply a list of
synsets in the ImageNet Challenge and output the list of bounding boxes
associated with the training images of the ILSVRC.
We use these bounding boxes to inform the random distortion of images
supplied to the network.
If you run this script successfully, you will see the following output
to stderr:
> Finished processing 544546 XML files.
> Skipped 0 XML files not in ImageNet Challenge.
> Skipped 0 bounding boxes not in ImageNet Challenge.
> Wrote 615299 bounding boxes from 544546 annotated images.
"""
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
import glob
import os.path
import sys
import xml.etree.ElementTree as ET
class BoundingBox(object):
pass
def GetItem(name, root, index=0):
count = 0
for item in root.iter(name):
if count == index:
return item.text
count += 1
# Failed to find "index" occurrence of item.
return -1
def GetInt(name, root, index=0):
return int(GetItem(name, root, index))
def FindNumberBoundingBoxes(root):
index = 0
while True:
if GetInt('xmin', root, index) == -1:
break
index += 1
return index
def ProcessXMLAnnotation(xml_file):
"""Process a single XML file containing a bounding box."""
# pylint: disable=broad-except
try:
tree = ET.parse(xml_file)
except Exception:
print('Failed to parse: ' + xml_file, file=sys.stderr)
return None
# pylint: enable=broad-except
root = tree.getroot()
num_boxes = FindNumberBoundingBoxes(root)
boxes = []
for index in xrange(num_boxes):
box = BoundingBox()
# Grab the 'index' annotation.
box.xmin = GetInt('xmin', root, index)
box.ymin = GetInt('ymin', root, index)
box.xmax = GetInt('xmax', root, index)
box.ymax = GetInt('ymax', root, index)
box.width = GetInt('width', root)
box.height = GetInt('height', root)
box.filename = GetItem('filename', root) + '.JPEG'
box.label = GetItem('name', root)
xmin = float(box.xmin) / float(box.width)
xmax = float(box.xmax) / float(box.width)
ymin = float(box.ymin) / float(box.height)
ymax = float(box.ymax) / float(box.height)
# Some images contain bounding box annotations that
# extend outside of the supplied image. See, e.g.
# n03127925/n03127925_147.xml
# Additionally, for some bounding boxes, the min > max
# or the box is entirely outside of the image.
min_x = min(xmin, xmax)
max_x = max(xmin, xmax)
box.xmin_scaled = min(max(min_x, 0.0), 1.0)
box.xmax_scaled = min(max(max_x, 0.0), 1.0)
min_y = min(ymin, ymax)
max_y = max(ymin, ymax)
box.ymin_scaled = min(max(min_y, 0.0), 1.0)
box.ymax_scaled = min(max(max_y, 0.0), 1.0)
boxes.append(box)
return boxes
if __name__ == '__main__':
if len(sys.argv) < 2 or len(sys.argv) > 3:
print('Invalid usage\n'
'usage: process_bounding_boxes.py <dir> [synsets-file]',
file=sys.stderr)
sys.exit(-1)
xml_files = glob.glob(sys.argv[1] + '/*/*.xml')
print('Identified %d XML files in %s' % (len(xml_files), sys.argv[1]),
file=sys.stderr)
if len(sys.argv) == 3:
labels = set([l.strip() for l in open(sys.argv[2]).readlines()])
print('Identified %d synset IDs in %s' % (len(labels), sys.argv[2]),
file=sys.stderr)
else:
labels = None
skipped_boxes = 0
skipped_files = 0
saved_boxes = 0
saved_files = 0
for file_index, one_file in enumerate(xml_files):
# Example: <...>/n06470073/n00141669_6790.xml
label = os.path.basename(os.path.dirname(one_file))
# Determine if the annotation is from an ImageNet Challenge label.
if labels is not None and label not in labels:
skipped_files += 1
continue
bboxes = ProcessXMLAnnotation(one_file)
assert bboxes is not None, 'No bounding boxes found in ' + one_file
found_box = False
for bbox in bboxes:
if labels is not None:
if bbox.label != label:
# Note: There is a slight bug in the bounding box annotation data.
# Many of the dog labels have the human label 'Scottish_deerhound'
# instead of the synset ID 'n02092002' in the bbox.label field. As a
# simple hack to overcome this issue, we only exclude bbox labels
# *which are synset ID's* that do not match original synset label for
# the XML file.
if bbox.label in labels:
skipped_boxes += 1
continue
# Guard against improperly specified boxes.
if (bbox.xmin_scaled >= bbox.xmax_scaled or
bbox.ymin_scaled >= bbox.ymax_scaled):
skipped_boxes += 1
continue
# Note bbox.filename occasionally contains '%s' in the name. This is
# data set noise that is fixed by just using the basename of the XML file.
image_filename = os.path.splitext(os.path.basename(one_file))[0]
print('%s.JPEG,%.4f,%.4f,%.4f,%.4f' %
(image_filename,
bbox.xmin_scaled, bbox.ymin_scaled,
bbox.xmax_scaled, bbox.ymax_scaled))
saved_boxes += 1
found_box = True
if found_box:
saved_files += 1
else:
skipped_files += 1
if not file_index % 5000:
print('--> processed %d of %d XML files.' %
(file_index + 1, len(xml_files)),
file=sys.stderr)
print('--> skipped %d boxes and %d XML files.' %
(skipped_boxes, skipped_files), file=sys.stderr)
print('Finished processing %d XML files.' % len(xml_files), file=sys.stderr)
print('Skipped %d XML files not in ImageNet Challenge.' % skipped_files,
file=sys.stderr)
print('Skipped %d bounding boxes not in ImageNet Challenge.' % skipped_boxes,
file=sys.stderr)
print('Wrote %d bounding boxes from %d annotated images.' %
(saved_boxes, saved_files),
file=sys.stderr)
print('Finished.', file=sys.stderr)
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment