weka.dl4j.interpretability

Class WekaScoreCAM

java.lang.Object

weka.dl4j.interpretability.AbstractCNNSaliencyMapWrapper

weka.dl4j.interpretability.WekaScoreCAM

All Implemented Interfaces:

java.io.Serializable, weka.core.OptionHandler

public class WekaScoreCAM
extends AbstractCNNSaliencyMapWrapper

WEKA Wrapper for the Deeplearning4j ScoreCAM implementation. Valid options are:

 -bs <int>
  The mini batch size to use for map generation

 -target-classes <int,int,...>
  Output class to generate saliency maps for; default is -1 (use the highest probability class). This only needs to be set if wanting to use a non-default class from the *command line*; if using the *GUI*, the 'View Saliency Map' window contains the interface for setting this.

 -output <file location>
  File for the saliency map to be saved in

 -normalize
  When generating the heatmap, should the values be normalized to be in [0, 1]

See Also:

Serialized Form

Constructor Summary

Constructors

Constructor and Description
WekaScoreCAM()

Method Summary

Modifier and Type	Method and Description
java.awt.image.BufferedImage	generateHeatmapToImage() Generates heatmaps, returning a human-viewable heatmap summary.
void	processImage(java.io.File imageFile) Main processing entrypoint.

Methods inherited from class weka.dl4j.interpretability.AbstractCNNSaliencyMapWrapper

getBatchSize, getClassMap, getCustomModelSetup, getDl4jMlpClassifier, getNormalizeHeatmap, getOptions, getOutputFile, getProgressManager, getTargetClassIDs, getTargetClassIDsAsInt, listOptions, saveResult, setBatchSize, setClassMap, setCustomModelSetup, setDl4jMlpClassifier, setNormalizeHeatmap, setOptions, setOutputFile, setTargetClassIDs, setTargetClassIDsAsInt

Methods inherited from class java.lang.Object

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

Methods inherited from interface weka.core.OptionHandler

makeCopy

Constructor Detail

WekaScoreCAM

public WekaScoreCAM()

Method Detail

processImage

public void processImage(java.io.File imageFile)

Description copied from class: AbstractCNNSaliencyMapWrapper

Main processing entrypoint. Takes an image as input, and performs the necessary processing on it. Note that this is performed separately to the actual saliency map generation. This method should only perform the expensive one-time processing on an image.

Specified by:

processImage in class AbstractCNNSaliencyMapWrapper

Parameters:

imageFile - Image to process heatmap for.

generateHeatmapToImage

public java.awt.image.BufferedImage generateHeatmapToImage()

Description copied from class: AbstractCNNSaliencyMapWrapper

Generates heatmaps, returning a human-viewable heatmap summary.

Specified by:

generateHeatmapToImage in class AbstractCNNSaliencyMapWrapper

Returns:

Human-viewable heatmap summary.