def test_image_decoder(self):
# Make random image.
image_size = 84
image = np.random.randint(low=0,
high=255,
size=[image_size, image_size,
3]).astype(np.ubyte)
# Encode
image_bytes = dataset_to_records.encode_image(image,
image_format='PNG')
label = np.zeros(1).astype(np.int64)
image_example = dataset_to_records.make_example(image_bytes,
label,
input_key='image',
label_key='label')
# Decode
image_decoder = decoder.ImageDecoder(image_size=image_size)
image_decoded = image_decoder(image_example)
# Assert perfect reconstruction.
with self.session(use_gpu=False) as sess:
image_rec_numpy = sess.run(image_decoded)
self.assertAllClose(2 * (image.astype(np.float32) / 255.0 - 0.5),
image_rec_numpy)
def write_feature_records(features, label, output_path):
"""Creates a record file from features and labels.
Args:
features: An [n, m] numpy array of features.
label: An integer, the label common to all records.
output_path: A string specifying the location of the record.
"""
writer = tf.python_io.TFRecordWriter(output_path)
for feat in list(features):
# Write the example.
serialized_example = dataset_to_records.make_example([
('image/embedding', 'float32', feat.tolist()),
('image/class/label', 'int64', [label])
])
writer.write(serialized_example)
writer.close()
def test_feature_decoder(self):
# Make random feature.
feat_size = 64
feat = np.random.randn(feat_size).astype(np.float32)
label = np.zeros(1).astype(np.int64)
# Encode
feat_example = dataset_to_records.make_example([
('image/embedding', 'float32', feat),
('image/class/label', 'int64', [label]),
])
# Decode
feat_decoder = decoder.FeatureDecoder(feat_len=feat_size)
feat_decoded = feat_decoder(feat_example)
# Assert perfect reconstruction.
with self.session(use_gpu=False) as sess:
feat_rec_numpy = sess.run(feat_decoded)
self.assertAllEqual(feat_rec_numpy, feat)
def test_feature_decoder(self):
# Make random feature.
feat_size = 64
feat = np.random.randn(feat_size).astype(np.float32)
label = np.zeros(1).astype(np.int64)
with self.session(use_gpu=False) as sess:
feat_serial = sess.run(tf.io.serialize_tensor(feat))
# Encode
feat_example = dataset_to_records.make_example(
feat_serial,
label,
input_key='image/embedding',
label_key='image/class/label')
# Decode
feat_decoder = decoder.FeatureDecoder()
feat_decoded = feat_decoder(feat_example)
# Assert perfect reconstruction.
with self.session(use_gpu=False) as sess:
feat_rec_numpy = sess.run(feat_decoded)
self.assertAllEqual(feat_rec_numpy, feat)
def test_string_decoder(self):
# Make random image.
image_size = 32
image = np.random.randint(low=0,
high=255,
size=[image_size, image_size,
3]).astype(np.ubyte)
# Encode
image_bytes = dataset_to_records.encode_image(image,
image_format='PNG')
label = np.zeros(1).astype(np.int64)
image_example = dataset_to_records.make_example([
('image', 'bytes', [image_bytes]), ('label', 'int64', [label])
])
# Decode
string_decoder = decoder.StringDecoder()
image_string = string_decoder(image_example)
decoded_image = tf.image.decode_image(image_string)
# Assert perfect reconstruction.
with self.session(use_gpu=False) as sess:
image_decoded = sess.run(decoded_image)
self.assertAllClose(image, image_decoded)
