Github user MrBago commented on a diff in the pull request:
https://github.com/apache/spark/pull/19439#discussion_r145843289
--- Diff: python/pyspark/ml/image.py ---
@@ -0,0 +1,122 @@
+#
+# Licensed to the Apache Software Foundation (ASF) under one or more
+# contributor license agreements. See the NOTICE file distributed with
+# this work for additional information regarding copyright ownership.
+# The ASF licenses this file to You 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.
+#
+
+import pyspark
+from pyspark import SparkContext
+from pyspark.sql.types import *
+from pyspark.sql.types import Row, _create_row
+from pyspark.sql import DataFrame
+from pyspark.ml.param.shared import *
+import numpy as np
+
+undefinedImageType = "Undefined"
+
+imageFields = ["origin", "height", "width", "nChannels", "mode", "data"]
+
+ocvTypes = {
+ undefinedImageType: -1,
+ "CV_8U": 0, "CV_8UC1": 0, "CV_8UC2": 8, "CV_8UC3": 16, "CV_8UC4": 24,
+ "CV_8S": 1, "CV_8SC1": 1, "CV_8SC2": 9, "CV_8SC3": 17, "CV_8SC4": 25,
+ "CV_16U": 2, "CV_16UC1": 2, "CV_16UC2": 10, "CV_16UC3": 18,
"CV_16UC4": 26,
+ "CV_16S": 3, "CV_16SC1": 3, "CV_16SC2": 11, "CV_16SC3": 19,
"CV_16SC4": 27,
+ "CV_32S": 4, "CV_32SC1": 4, "CV_32SC2": 12, "CV_32SC3": 20,
"CV_32SC4": 28,
+ "CV_32F": 5, "CV_32FC1": 5, "CV_32FC2": 13, "CV_32FC3": 21,
"CV_32FC4": 29,
+ "CV_64F": 6, "CV_64FC1": 6, "CV_64FC2": 14, "CV_64FC3": 22,
"CV_64FC4": 30
+}
+
+# DataFrame with a single column of images named "image" (nullable)
+imageSchema = StructType(StructField("image", StructType([
+ StructField(imageFields[0], StringType(), True),
+ StructField(imageFields[1], IntegerType(), False),
+ StructField(imageFields[2], IntegerType(), False),
+ StructField(imageFields[3], IntegerType(), False),
+ # OpenCV-compatible type: CV_8UC3 in most cases
+ StructField(imageFields[4], StringType(), False),
+ # bytes in OpenCV-compatible order: row-wise BGR in most cases
+ StructField(imageFields[5], BinaryType(), False)]), True))
+
+
+def toNDArray(image):
+ """
+ Converts an image to a one-dimensional array.
+
+ :param image (object): The image to be converted
+ :rtype array: The image as a one-dimensional array
+
+ .. versionadded:: 2.3.0
+ """
+ height = image.height
+ width = image.width
+ nChannels = image.nChannels
+ return np.ndarray(
+ shape=(height, width, nChannels),
+ dtype=np.uint8,
+ buffer=image.data,
+ strides=(width * nChannels, nChannels, 1))
+
+
+def toImage(array, origin="", mode=ocvTypes["CV_8UC3"]):
--- End diff --
The interaction between mode and dtype here is unclear here. I think there
are two sensible things to do.
1) determine the dtype from the mode, and use that to cast the image, ie:
`toImage(a, mode="CV_8UC3") =>
bytearray(array.astype(dtype=np.uint8).ravel())`
`toImage(a, mode="CV_32FC3") =>
bytearray(array.astype(dtype=np.float32).ravel())`
2) Remove the mode argument to the function and use `array.dtype` to infer
the mode.
Also we don't currently check that nChannels is consistent with the mode
proved. A (H, W, 1) image will currently get assigned mode "CV_8UC3".
I think it's fine to restrict the set of modes/dtypes we support but again
we should be careful not to return a malformed image row.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
