Package org.datavec.image.recordreader.objdetect

Class ObjectDetectionRecordReader

java.lang.Object

org.datavec.api.records.reader.BaseRecordReader

org.datavec.image.recordreader.BaseImageRecordReader

org.datavec.image.recordreader.objdetect.ObjectDetectionRecordReader

All Implemented Interfaces:

Closeable, Serializable, AutoCloseable, Configurable, RecordReader

public class ObjectDetectionRecordReader
extends BaseImageRecordReader

See Also:

Serialized Form

Field Summary

Fields

Modifier and Type	Field	Description
protected Image	currentImage

Fields inherited from class org.datavec.image.recordreader.BaseImageRecordReader

appendLabel, channels, CHANNELS, conf, CROP_IMAGE, cropImage, currentFile, fileNameMap, finishedInputStreamSplit, height, HEIGHT, hitImage, IMAGE_LOADER, imageLoader, imageTransform, inputSplit, iter, labelGenerator, labelMultiGenerator, labels, logLabelCountOnInit, nchw_channels_first, pattern, patternPosition, record, width, WIDTH, writeLabel

Fields inherited from class org.datavec.api.records.reader.BaseRecordReader

listeners, streamCreatorFn

Fields inherited from interface org.datavec.api.records.reader.RecordReader

APPEND_LABEL, LABELS, NAME_SPACE

Constructor Summary

Constructors

Constructor	Description
ObjectDetectionRecordReader(int height, int width, int channels, int gridH, int gridW, boolean nchw, ImageObjectLabelProvider labelProvider)	Create ObjectDetectionRecordReader with
ObjectDetectionRecordReader(int height, int width, int channels, int gridH, int gridW, boolean nchw, ImageObjectLabelProvider labelProvider, ImageTransform imageTransform)	When imageTransform != null, object is removed if new center is outside of transformed image bounds.
ObjectDetectionRecordReader(int height, int width, int channels, int gridH, int gridW, ImageObjectLabelProvider labelProvider)	As per ObjectDetectionRecordReader(int, int, int, int, int, boolean, ImageObjectLabelProvider) but hardcoded to NCHW format
ObjectDetectionRecordReader(int height, int width, int channels, int gridH, int gridW, ImageObjectLabelProvider labelProvider, ImageTransform imageTransform)	As per ObjectDetectionRecordReader(int, int, int, int, int, boolean, ImageObjectLabelProvider, ImageTransform) but hardcoded to NCHW format

Method Summary

Modifier and Type	Method	Description
void	initialize(InputSplit split)
List<Writable>	next()
List<List<Writable>>	next(int num)
Record	nextRecord()
List<Writable>	record(URI uri, DataInputStream dataInputStream)

Methods inherited from class org.datavec.image.recordreader.BaseImageRecordReader

accumulateLabel, batchesSupported, close, containsFormat, getConf, getCurrentFile, getLabel, getLabels, hasNext, initialize, initialize, initialize, loadFromMetaData, loadFromMetaData, numLabels, reset, resetSupported, setConf, setCurrentFile, setLabels

Methods inherited from class org.datavec.api.records.reader.BaseRecordReader

getListeners, invokeListeners, setListeners, setListeners

Methods inherited from class java.lang.Object

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

Field Detail

currentImage

protected Image currentImage

Constructor Detail

ObjectDetectionRecordReader

public ObjectDetectionRecordReader(int height,
                                   int width,
                                   int channels,
                                   int gridH,
                                   int gridW,
                                   ImageObjectLabelProvider labelProvider)

As per ObjectDetectionRecordReader(int, int, int, int, int, boolean, ImageObjectLabelProvider) but hardcoded to NCHW format

ObjectDetectionRecordReader

public ObjectDetectionRecordReader(int height,
                                   int width,
                                   int channels,
                                   int gridH,
                                   int gridW,
                                   boolean nchw,
                                   ImageObjectLabelProvider labelProvider)

Create ObjectDetectionRecordReader with

Parameters:

height - Height of the output images

width - Width of the output images

channels - Number of channels for the output images

gridH - Grid/quantization size (along height dimension) - Y axis

gridW - Grid/quantization size (along height dimension) - X axis

nchw - If true: return NCHW format labels with array shape [minibatch, 4+C, h, w]; if false, return NHWC format labels with array shape [minibatch, h, w, 4+C]

labelProvider - ImageObjectLabelProvider - used to look up which objects are in each image

ObjectDetectionRecordReader

public ObjectDetectionRecordReader(int height,
                                   int width,
                                   int channels,
                                   int gridH,
                                   int gridW,
                                   ImageObjectLabelProvider labelProvider,
                                   ImageTransform imageTransform)

As per ObjectDetectionRecordReader(int, int, int, int, int, boolean, ImageObjectLabelProvider, ImageTransform) but hardcoded to NCHW format

ObjectDetectionRecordReader

public ObjectDetectionRecordReader(int height,
                                   int width,
                                   int channels,
                                   int gridH,
                                   int gridW,
                                   boolean nchw,
                                   ImageObjectLabelProvider labelProvider,
                                   ImageTransform imageTransform)

When imageTransform != null, object is removed if new center is outside of transformed image bounds.

Parameters:

height - Height of the output images

width - Width of the output images

channels - Number of channels for the output images

gridH - Grid/quantization size (along height dimension) - Y axis

gridW - Grid/quantization size (along height dimension) - X axis

labelProvider - ImageObjectLabelProvider - used to look up which objects are in each image

nchw - If true: return NCHW format labels with array shape [minibatch, 4+C, h, w]; if false, return NHWC format labels with array shape [minibatch, h, w, 4+C]

imageTransform - ImageTransform - used to transform image and coordinates

Method Detail

next

public List<Writable> next()

Specified by:

next in interface RecordReader

Overrides:

next in class BaseImageRecordReader

initialize

public void initialize(InputSplit split)
                throws IOException

Specified by:

initialize in interface RecordReader

Overrides:

initialize in class BaseImageRecordReader

Throws:

IOException

next

public List<List<Writable>> next(int num)

Specified by:

next in interface RecordReader

Overrides:

next in class BaseImageRecordReader

record

public List<Writable> record(URI uri,
                             DataInputStream dataInputStream)
                      throws IOException

Specified by:

record in interface RecordReader

Overrides:

record in class BaseImageRecordReader

Throws:

IOException

nextRecord

public Record nextRecord()

Specified by:

nextRecord in interface RecordReader

Overrides:

nextRecord in class BaseImageRecordReader