def test_prediction_vs_tensorflow_inceptionV3(self):
output_col = "prediction"
image_df = image_utils.getSampleImageDF()
# An example of how a pre-trained keras model can be used with TFImageTransformer
with KSessionWrap() as (sess, g):
with g.as_default():
K.set_learning_phase(0) # this is important but it's on the user to call it.
# nChannels needed for input_tensor in the InceptionV3 call below
image_string = utils.imageInputPlaceholder(nChannels=3)
resized_images = tf.image.resize_images(image_string,
InceptionV3Constants.INPUT_SHAPE)
# keras expects array in RGB order, we get it from image schema in BGR =>
# need to flip
preprocessed = preprocess_input(imageIO._reverseChannels(resized_images))
model = InceptionV3(input_tensor=preprocessed, weights="imagenet")
graph = tfx.strip_and_freeze_until([model.output], g, sess, return_graph=True)
transformer = TFImageTransformer(channelOrder='BGR', inputCol="image", outputCol=output_col, graph=graph,
inputTensor=image_string, outputTensor=model.output,
outputMode="vector")
transformed_df = transformer.transform(image_df.limit(10))
self.assertDfHasCols(transformed_df, [output_col])
collected = transformed_df.collect()
transformer_values, transformer_topK = self.transformOutputToComparables(collected,
output_col, lambda row: row['image']['origin'])
tf_values, tf_topK = self._executeTensorflow(graph, image_string.name, model.output.name,
image_df)
self.compareClassSets(tf_topK, transformer_topK)
self.compareClassOrderings(tf_topK, transformer_topK)
self.compareArrays(tf_values, transformer_values, decimal=5)
def test_prediction_vs_tensorflow_inceptionV3(self):
output_col = "prediction"
image_df = image_utils.getSampleImageDF()
# An example of how a pre-trained keras model can be used with TFImageTransformer
with KSessionWrap() as (sess, g):
with g.as_default():
K.set_learning_phase(0) # this is important but it's on the user to call it.
# nChannels needed for input_tensor in the InceptionV3 call below
image_string = utils.imageInputPlaceholder(nChannels=3)
resized_images = tf.image.resize_images(image_string,
InceptionV3Constants.INPUT_SHAPE)
# keras expects array in RGB order, we get it from image schema in BGR => need to flip
preprocessed = preprocess_input(imageIO._reverseChannels(resized_images))
model = InceptionV3(input_tensor=preprocessed, weights="imagenet")
graph = tfx.strip_and_freeze_until([model.output], g, sess, return_graph=True)
transformer = TFImageTransformer(channelOrder='BGR', inputCol="image", outputCol=output_col, graph=graph,
inputTensor=image_string, outputTensor=model.output,
outputMode="vector")
transformed_df = transformer.transform(image_df.limit(10))
self.assertDfHasCols(transformed_df, [output_col])
collected = transformed_df.collect()
transformer_values, transformer_topK = self.transformOutputToComparables(collected,
output_col, lambda row: row['image']['origin'])
tf_values, tf_topK = self._executeTensorflow(graph, image_string.name, model.output.name,
image_df)
self.compareClassSets(tf_topK, transformer_topK)
self.compareClassOrderings(tf_topK, transformer_topK)
self.compareArrays(tf_values, transformer_values)
def test_load_image_vs_keras(self):
g = tf.Graph()
with g.as_default():
image_arr = utils.imageInputPlaceholder()
# keras expects array in RGB order, we get it from image schema in BGR => need to flip
preprocessed = preprocess_input(
imageIO._reverseChannels(image_arr))
output_col = "transformed_image"
transformer = TFImageTransformer(channelOrder='BGR',
inputCol="image",
outputCol=output_col,
graph=g,
inputTensor=image_arr,
outputTensor=preprocessed.name,
outputMode="vector")
image_df = image_utils.getSampleImageDF()
df = transformer.transform(image_df.limit(5))
for row in df.collect():
processed = np.array(row[output_col]).astype(np.float32)
# compare to keras loading
images = self._loadImageViaKeras(row["image"]['origin'])
image = images[0]
image.shape = (1, image.shape[0] * image.shape[1] * image.shape[2])
keras_processed = image[0]
self.assertTrue((processed == keras_processed).all())
def test_resize(self):
self.assertRaises(ValueError, imageIO.createResizeImageUDF, [1, 2, 3])
make_smaller = imageIO.createResizeImageUDF([4, 5]).func
imgAsRow = imageIO.imageArrayToStruct(array)
smallerImg = make_smaller(imgAsRow)
self.assertEqual(smallerImg.height, 4)
self.assertEqual(smallerImg.width, 5)
# Compare to PIL resizing
imgAsPIL = PIL.Image.fromarray(obj=imageIO._reverseChannels(array)).resize((5, 4))
smallerAry = imageIO._reverseChannels(np.asarray(imgAsPIL))
np.testing.assert_array_equal(smallerAry, imageIO.imageStructToArray(smallerImg))
# Test that resize with the same size is a no-op
sameImage = imageIO.createResizeImageUDF((imgAsRow.height, imgAsRow.width)).func(imgAsRow)
self.assertEqual(imgAsRow, sameImage)
# Test that we have a valid image schema (all fields are in)
for n in ImageSchema.imageSchema['image'].dataType.names:
smallerImg[n]
def test_resize(self):
self.assertRaises(ValueError, imageIO.createResizeImageUDF, [1, 2, 3])
make_smaller = imageIO.createResizeImageUDF([4, 5]).func
imgAsRow = imageIO.imageArrayToStruct(array)
smallerImg = make_smaller(imgAsRow)
self.assertEqual(smallerImg.height, 4)
self.assertEqual(smallerImg.width, 5)
# Compare to PIL resizing
imgAsPIL = PIL.Image.fromarray(
obj=imageIO._reverseChannels(array)).resize((5, 4))
smallerAry = imageIO._reverseChannels(np.asarray(imgAsPIL))
np.testing.assert_array_equal(smallerAry,
imageIO.imageStructToArray(smallerImg))
# Test that resize with the same size is a no-op
sameImage = imageIO.createResizeImageUDF(
(imgAsRow.height, imgAsRow.width)).func(imgAsRow)
self.assertEqual(imgAsRow, sameImage)
# Test that we have a valid image schema (all fields are in)
for n in ImageSchema.imageSchema['image'].dataType.names:
smallerImg[n]
def create_image_data():
# Random image-like data
array = np.random.randint(0, 256, (10, 11, 3), 'uint8')
# Compress as png
imgFile = BytesIO()
PIL.Image.fromarray(array).save(imgFile, 'png')
imgFile.seek(0)
# Get Png data as stream
pngData = imgFile.read()
# PIL is RGB but image schema is BGR => flip the channels
return imageIO._reverseChannels(array), pngData
def create_image_data():
# Random image-like data
array = np.random.randint(0, 256, (10, 11, 3), 'uint8')
# Compress as png
imgFile = BytesIO()
PIL.Image.fromarray(array).save(imgFile, 'png')
imgFile.seek(0)
# Get Png data as stream
pngData = imgFile.read()
# PIL is RGB but image schema is BGR => flip the channels
return imageIO._reverseChannels(array), pngData
def _preprocessingInceptionV3Transformed(self, outputMode, outputCol):
g = tf.Graph()
with g.as_default():
image_arr = utils.imageInputPlaceholder()
resized_images = tf.image.resize_images(image_arr, InceptionV3Constants.INPUT_SHAPE)
# keras expects array in RGB order, we get it from image schema in BGR => need to flip
processed_images = preprocess_input(imageIO._reverseChannels(resized_images))
self.assertEqual(processed_images.shape[1], InceptionV3Constants.INPUT_SHAPE[0])
self.assertEqual(processed_images.shape[2], InceptionV3Constants.INPUT_SHAPE[1])
transformer = TFImageTransformer(channelOrder='BGR', inputCol="image", outputCol=outputCol, graph=g,
inputTensor=image_arr.name, outputTensor=processed_images,
outputMode=outputMode)
image_df = image_utils.getSampleImageDF()
return transformer.transform(image_df.limit(5))
def _preprocessingInceptionV3Transformed(self, outputMode, outputCol):
g = tf.Graph()
with g.as_default():
image_arr = utils.imageInputPlaceholder()
resized_images = tf.image.resize_images(image_arr, InceptionV3Constants.INPUT_SHAPE)
# keras expects array in RGB order, we get it from image schema in BGR => need to flip
processed_images = preprocess_input(imageIO._reverseChannels(resized_images))
self.assertEqual(processed_images.shape[1], InceptionV3Constants.INPUT_SHAPE[0])
self.assertEqual(processed_images.shape[2], InceptionV3Constants.INPUT_SHAPE[1])
transformer = TFImageTransformer(channelOrder='BGR', inputCol="image", outputCol=outputCol, graph=g,
inputTensor=image_arr.name, outputTensor=processed_images,
outputMode=outputMode)
image_df = image_utils.getSampleImageDF()
return transformer.transform(image_df.limit(5))
def test_load_image_vs_keras(self):
g = tf.Graph()
with g.as_default():
image_arr = utils.imageInputPlaceholder()
# keras expects array in RGB order, we get it from image schema in BGR => need to flip
preprocessed = preprocess_input(imageIO._reverseChannels(image_arr))
output_col = "transformed_image"
transformer = TFImageTransformer(channelOrder='BGR', inputCol="image", outputCol=output_col, graph=g,
inputTensor=image_arr, outputTensor=preprocessed.name,
outputMode="vector")
image_df = image_utils.getSampleImageDF()
df = transformer.transform(image_df.limit(5))
for row in df.collect():
processed = np.array(row[output_col]).astype(np.float32)
# compare to keras loading
images = self._loadImageViaKeras(row["image"]['origin'])
image = images[0]
image.shape = (1, image.shape[0] * image.shape[1] * image.shape[2])
keras_processed = image[0]
np.testing.assert_array_almost_equal(keras_processed, processed, decimal=6)
def load_image_uri_impl(uri):
try:
return imageArrayToStruct(_reverseChannels(loader(uri)))
except BaseException: # pylint: disable=bare-except
return None
def pil_load_spimg(fpath):
from PIL import Image
import numpy as np
img_arr = np.array(Image.open(fpath), dtype=np.uint8)
# PIL is RGB, image schema is BGR => need to flip the channels
return imageArrayToStruct(_reverseChannels(img_arr))
def keras_load_spimg(fpath):
# Keras loads image in RGB order, ImageSchema expects BGR => need to flip
return imageArrayToStruct(_reverseChannels(keras_load_img(fpath)))
def preprocess(self, inputImage):
# Keras expects RGB order
return xception.preprocess_input(_reverseChannels(inputImage))
def load_image_uri_impl(uri):
try:
return imageArrayToStruct(_reverseChannels(loader(uri)))
except BaseException: # pylint: disable=bare-except
return None
def preprocess(self, inputImage):
# Keras expects RGB order
return xception.preprocess_input(_reverseChannels(inputImage))
