forked from antoine77340/Youtube-8M-WILLOW
/
inference.py
203 lines (169 loc) · 8.03 KB
/
inference.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
# 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.
"""Binary for generating predictions over a set of videos."""
import os
import time
import numpy
import tensorflow as tf
from tensorflow import app
from tensorflow import flags
from tensorflow import gfile
from tensorflow import logging
import eval_util
import losses
import readers
import utils
FLAGS = flags.FLAGS
if __name__ == '__main__':
flags.DEFINE_string("train_dir", "/tmp/yt8m_model/",
"The directory to load the model files from.")
flags.DEFINE_string("output_file", "",
"The file to save the predictions to.")
flags.DEFINE_string(
"input_data_pattern", "",
"File glob defining the evaluation dataset in tensorflow.SequenceExample "
"format. The SequenceExamples are expected to have an 'rgb' byte array "
"sequence feature as well as a 'labels' int64 context feature.")
# Model flags.
flags.DEFINE_bool(
"frame_features", False,
"If set, then --eval_data_pattern must be frame-level features. "
"Otherwise, --eval_data_pattern must be aggregated video-level "
"features. The model must also be set appropriately (i.e. to read 3D "
"batches VS 4D batches.")
flags.DEFINE_integer(
"batch_size", 8192,
"How many examples to process per batch.")
flags.DEFINE_string("feature_names", "mean_rgb", "Name of the feature "
"to use for training.")
flags.DEFINE_string("feature_sizes", "1024", "Length of the feature vectors.")
# Other flags.
flags.DEFINE_integer("num_readers", 4,
"How many threads to use for reading input files.")
flags.DEFINE_integer("top_k", 20,
"How many predictions to output per video.")
flags.DEFINE_integer("check_point",-1,
"Model checkpoint to load, -1 for latest.")
def format_lines(video_ids, predictions, top_k):
batch_size = len(video_ids)
for video_index in range(batch_size):
top_indices = numpy.argpartition(predictions[video_index], -top_k)[-top_k:]
line = [(class_index, predictions[video_index][class_index])
for class_index in top_indices]
line = sorted(line, key=lambda p: -p[1])
yield video_ids[video_index].decode('utf-8') + "," + " ".join("%i %f" % pair
for pair in line) + "\n"
def get_input_data_tensors(reader, data_pattern, batch_size, num_readers=1):
"""Creates the section of the graph which reads the input data.
Args:
reader: A class which parses the input data.
data_pattern: A 'glob' style path to the data files.
batch_size: How many examples to process at a time.
num_readers: How many I/O threads to use.
Returns:
A tuple containing the features tensor, labels tensor, and optionally a
tensor containing the number of frames per video. The exact dimensions
depend on the reader being used.
Raises:
IOError: If no files matching the given pattern were found.
"""
with tf.name_scope("input"):
files = gfile.Glob(data_pattern)
if not files:
raise IOError("Unable to find input files. data_pattern='" +
data_pattern + "'")
logging.info("number of input files: " + str(len(files)))
filename_queue = tf.train.string_input_producer(
files, num_epochs=1, shuffle=False)
examples_and_labels = [reader.prepare_reader(filename_queue)
for _ in range(num_readers)]
video_id_batch, video_batch, unused_labels, num_frames_batch = (
tf.train.batch_join(examples_and_labels,
batch_size=batch_size,
allow_smaller_final_batch = True,
enqueue_many=True))
return video_id_batch, video_batch, num_frames_batch
def inference(reader, train_dir, data_pattern, out_file_location, batch_size, top_k):
with tf.Session() as sess, gfile.Open(out_file_location, "w+") as out_file:
video_id_batch, video_batch, num_frames_batch = get_input_data_tensors(reader, data_pattern, batch_size)
latest_checkpoint = tf.train.latest_checkpoint(train_dir)
if latest_checkpoint is None:
raise Exception("unable to find a checkpoint at location: %s" % train_dir)
else:
if FLAGS.check_point < 0:
meta_graph_location = latest_checkpoint + ".meta"
else:
meta_graph_location = FLAGS.train_dir + "/model.ckpt-" + str(FLAGS.check_point) + ".meta"
latest_checkpoint = FLAGS.train_dir + "/model.ckpt-" + str(FLAGS.check_point)
logging.info("loading meta-graph: " + meta_graph_location)
saver = tf.train.import_meta_graph(meta_graph_location, clear_devices=True)
logging.info("restoring variables from " + latest_checkpoint)
saver.restore(sess, latest_checkpoint)
input_tensor = tf.get_collection("input_batch_raw")[0]
num_frames_tensor = tf.get_collection("num_frames")[0]
predictions_tensor = tf.get_collection("predictions")[0]
# Workaround for num_epochs issue.
def set_up_init_ops(variables):
init_op_list = []
for variable in list(variables):
if "train_input" in variable.name:
init_op_list.append(tf.assign(variable, 1))
variables.remove(variable)
init_op_list.append(tf.variables_initializer(variables))
return init_op_list
sess.run(set_up_init_ops(tf.get_collection_ref(
tf.GraphKeys.LOCAL_VARIABLES)))
coord = tf.train.Coordinator()
threads = tf.train.start_queue_runners(sess=sess, coord=coord)
num_examples_processed = 0
start_time = time.time()
out_file.write("VideoId,LabelConfidencePairs\n")
try:
while not coord.should_stop():
video_id_batch_val, video_batch_val,num_frames_batch_val = sess.run([video_id_batch, video_batch, num_frames_batch])
predictions_val, = sess.run([predictions_tensor], feed_dict={input_tensor: video_batch_val, num_frames_tensor: num_frames_batch_val})
now = time.time()
num_examples_processed += len(video_batch_val)
num_classes = predictions_val.shape[1]
logging.info("num examples processed: " + str(num_examples_processed) + " elapsed seconds: " + "{0:.2f}".format(now-start_time))
for line in format_lines(video_id_batch_val, predictions_val, top_k):
out_file.write(line)
out_file.flush()
except tf.errors.OutOfRangeError:
logging.info('Done with inference. The output file was written to ' + out_file_location)
finally:
coord.request_stop()
coord.join(threads)
sess.close()
def main(unused_argv):
logging.set_verbosity(tf.logging.INFO)
# convert feature_names and feature_sizes to lists of values
feature_names, feature_sizes = utils.GetListOfFeatureNamesAndSizes(
FLAGS.feature_names, FLAGS.feature_sizes)
if FLAGS.frame_features:
reader = readers.YT8MFrameFeatureReader(feature_names=feature_names,
feature_sizes=feature_sizes)
else:
reader = readers.YT8MAggregatedFeatureReader(feature_names=feature_names,
feature_sizes=feature_sizes)
if FLAGS.output_file is "":
raise ValueError("'output_file' was not specified. "
"Unable to continue with inference.")
if FLAGS.input_data_pattern is "":
raise ValueError("'input_data_pattern' was not specified. "
"Unable to continue with inference.")
inference(reader, FLAGS.train_dir, FLAGS.input_data_pattern,
FLAGS.output_file, FLAGS.batch_size, FLAGS.top_k)
if __name__ == "__main__":
app.run()