spark-reviews mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From MrBago <...@git.apache.org>
Subject [GitHub] spark pull request #19439: [SPARK-21866][ML][PySpark] Adding spark image rea...
Date Tue, 10 Oct 2017 17:35:11 GMT
Github user MrBago commented on a diff in the pull request:

    https://github.com/apache/spark/pull/19439#discussion_r143798662
  
    --- Diff: python/pyspark/ml/image.py ---
    @@ -0,0 +1,133 @@
    +#
    +# 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
    +}
    +
    +ImageSchema = 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)])
    +
    +
    +# TODO: generalize to other datatypes and number of channels
    +def toNDArray(image):
    +    """
    +    Converts an image to a 1-dimensional array
    +
    +    Args:
    +        image (object): The image to be converted
    +
    +    Returns:
    +        array: The image as a 1-dimensional array
    +
    +    .. versionadded:: 2.3.0
    +    """
    +    height = image.height
    +    width = image.width
    +    return np.asarray(image.data, dtype=np.uint8) \
    +             .reshape((height, width, 3))[:, :, (2, 1, 0)]
    --- End diff --
    
    This code assumes `image` is 3-channel GBR image and the user wants an RGB image. I think
we should try and support at least the 3 image types that `readImages` (CV_8UC1, CV_8UC3,
and CV_8UC4) but it would be nice to also support 1, 3, and 4 channel float images.
    
    The ndarray constructor is quite flexible and might be easier to work with than calling
`asarray` in this case because you want to treat the bytearray as a buffer not as a sequence
of ints.
    
    ```
    np.ndarray(
      shape=(height, width, nChannels),
      dtype=np.uint8,
      buffer=image.data,
      strides=(width * nChannels, nChannels, 1))
    ```
    
    Also I have mixed feelings about re-ordering the channels. I think it's probably useful
in the most common use-case, but (if my understanding is correct) the open cv types don't
require a specific channel order so we can't just assume the input is RGB or RGBA. Maybe we
should avoid re-ordering, document the ordering we use whenever appropriate, and then leave
it up to the user to do any necessary re-order for themselves.


---

---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org


Mime
View raw message