org.datavec.image.transform

Class LargestBlobCropTransform

java.lang.Object

org.datavec.image.transform.BaseImageTransform<org.bytedeco.opencv.opencv_core.Mat>

org.datavec.image.transform.LargestBlobCropTransform

All Implemented Interfaces:

ImageTransform

public class LargestBlobCropTransform
extends BaseImageTransform<org.bytedeco.opencv.opencv_core.Mat>

crop images based on it's largest blob. Calls internally opencv_imgproc.blur(Mat, Mat, Size) opencv_imgproc.Canny(Mat ,Mat, double, double) opencv_imgproc.threshold(Mat, Mat, double, double, int) opencv_imgproc.findContours(Mat, MatVector, Mat, int, int) opencv_imgproc.contourArea(Mat, boolean)

Author:

antdood

Field Summary

Fields

Modifier and Type	Field and Description
protected int	blurHeight
protected int	blurWidth
protected boolean	isCanny
protected int	lowerThresh
protected int	method
protected int	mode
protected Random	rng
protected int	upperThresh

Fields inherited from class org.datavec.image.transform.BaseImageTransform

converter, currentImage, random

Constructor Summary

Constructors

Constructor and Description
LargestBlobCropTransform() Calls this(null
LargestBlobCropTransform(Random random) Calls this(random, CV_RETR_CCOMP, CV_CHAIN_APPROX_SIMPLE, 3, 3, 100, 200, true)
LargestBlobCropTransform(Random random, int mode, int method, int blurWidth, int blurHeight, int lowerThresh, int upperThresh, boolean isCanny)

Method Summary

Modifier and Type	Method and Description
protected ImageWritable	doTransform(ImageWritable image, Random random) Takes an image and returns a cropped image based on it's largest blob.
float[]	query(float... coordinates) Transforms the given coordinates using the parameters that were used to transform the last image.

Methods inherited from class org.datavec.image.transform.BaseImageTransform

getCurrentImage, transform, transform

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Field Detail

rng

protected Random rng

mode

protected int mode

method

protected int method

blurWidth

protected int blurWidth

blurHeight

protected int blurHeight

upperThresh

protected int upperThresh

lowerThresh

protected int lowerThresh

isCanny

protected boolean isCanny

Constructor Detail

LargestBlobCropTransform

public LargestBlobCropTransform()

Calls this(null

LargestBlobCropTransform

public LargestBlobCropTransform(Random random)

Calls this(random, CV_RETR_CCOMP, CV_CHAIN_APPROX_SIMPLE, 3, 3, 100, 200, true)

LargestBlobCropTransform

public LargestBlobCropTransform(Random random,
                                int mode,
                                int method,
                                int blurWidth,
                                int blurHeight,
                                int lowerThresh,
                                int upperThresh,
                                boolean isCanny)

Parameters:

random - Object to use (or null for deterministic)

mode - Contour retrieval mode

method - Contour approximation method

blurWidth - Width of blurring kernel size

blurHeight - Height of blurring kernel size

lowerThresh - Lower threshold for either Canny or Threshold

upperThresh - Upper threshold for either Canny or Threshold

isCanny - Whether the edge detector is Canny or Threshold

Method Detail

doTransform

protected ImageWritable doTransform(ImageWritable image,
                                    Random random)

Takes an image and returns a cropped image based on it's largest blob.

Specified by:

doTransform in class BaseImageTransform<org.bytedeco.opencv.opencv_core.Mat>

Parameters:

image - to transform, null == end of stream

random - object to use (or null for deterministic)

Returns:

transformed image

query

public float[] query(float... coordinates)

Description copied from interface: ImageTransform

Transforms the given coordinates using the parameters that were used to transform the last image.

Specified by:

query in interface ImageTransform

Overrides:

query in class BaseImageTransform<org.bytedeco.opencv.opencv_core.Mat>

Parameters:

coordinates - to transforms (x1, y1, x2, y2, ...)

Returns:

transformed coordinates