-
Notifications
You must be signed in to change notification settings - Fork 5
/
SimpleCNN.java
170 lines (150 loc) · 7.61 KB
/
SimpleCNN.java
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
package org.genericsystem.cv.nn;
import java.io.File;
import java.io.IOException;
import java.util.Arrays;
import java.util.List;
import java.util.Random;
import org.datavec.api.io.filters.BalancedPathFilter;
import org.datavec.api.io.labels.ParentPathLabelGenerator;
import org.datavec.api.split.FileSplit;
import org.datavec.api.split.InputSplit;
import org.datavec.image.loader.BaseImageLoader;
import org.datavec.image.recordreader.ImageRecordReader;
import org.datavec.image.transform.FlipImageTransform;
import org.datavec.image.transform.ImageTransform;
import org.datavec.image.transform.WarpImageTransform;
import org.deeplearning4j.datasets.datavec.RecordReaderDataSetIterator;
import org.deeplearning4j.earlystopping.EarlyStoppingConfiguration;
import org.deeplearning4j.earlystopping.EarlyStoppingResult;
import org.deeplearning4j.earlystopping.scorecalc.DataSetLossCalculator;
import org.deeplearning4j.earlystopping.termination.MaxEpochsTerminationCondition;
import org.deeplearning4j.earlystopping.termination.ScoreImprovementEpochTerminationCondition;
import org.deeplearning4j.earlystopping.trainer.EarlyStoppingTrainer;
import org.deeplearning4j.eval.Evaluation;
import org.deeplearning4j.nn.api.OptimizationAlgorithm;
import org.deeplearning4j.nn.conf.GradientNormalization;
import org.deeplearning4j.nn.conf.MultiLayerConfiguration;
import org.deeplearning4j.nn.conf.NeuralNetConfiguration;
import org.deeplearning4j.nn.conf.Updater;
import org.deeplearning4j.nn.conf.inputs.InputType;
import org.deeplearning4j.nn.conf.layers.ConvolutionLayer;
import org.deeplearning4j.nn.conf.layers.DenseLayer;
import org.deeplearning4j.nn.conf.layers.OutputLayer;
import org.deeplearning4j.nn.conf.layers.SubsamplingLayer;
import org.deeplearning4j.nn.multilayer.MultiLayerNetwork;
import org.deeplearning4j.nn.weights.WeightInit;
import org.deeplearning4j.optimize.listeners.ScoreIterationListener;
import org.deeplearning4j.util.ModelSerializer;
import org.nd4j.linalg.activations.Activation;
import org.nd4j.linalg.dataset.api.iterator.DataSetIterator;
import org.nd4j.linalg.dataset.api.preprocessor.DataNormalization;
import org.nd4j.linalg.dataset.api.preprocessor.ImagePreProcessingScaler;
import org.nd4j.linalg.lossfunctions.LossFunctions;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;
public class SimpleCNN {
protected static final Logger log = LoggerFactory.getLogger(SimpleCNN.class);
private static final int seed = 123;
private static Random rng = new Random(seed);
private static final String[] allowedExtensions = BaseImageLoader.ALLOWED_FORMATS;
public static final Random randNumGen = new Random(seed);
private static int height = 250;
private static int width = 200;
private static int channels = 3;
protected static int iterations = 1;
public static void main(String[] args) throws Exception {
double learningRate = 0.005;
int batchSize = 4;
int nEpochs = 100;
File parentDir = new File(System.getProperty("user.dir"), "training");
FileSplit filesInDir = new FileSplit(parentDir, allowedExtensions, randNumGen);
ParentPathLabelGenerator labelMaker = new ParentPathLabelGenerator();
BalancedPathFilter pathFilter = new BalancedPathFilter(randNumGen, allowedExtensions, labelMaker, 0, 0, 100, 0);
InputSplit[] filesInDirSplit = filesInDir.sample(pathFilter, 70, 15, 15);
InputSplit trainData = filesInDirSplit[0];
InputSplit validData = filesInDirSplit[1];
InputSplit testData = filesInDirSplit[2];
ImageRecordReader recordReader = new ImageRecordReader(height, width, channels, labelMaker);
recordReader.initialize(trainData, null);
List<String> labels = recordReader.getLabels();
int outputNum = recordReader.numLabels();
MultiLayerConfiguration conf = new NeuralNetConfiguration.Builder()
.seed(seed)
.iterations(iterations)
.regularization(false)
.gradientNormalization(GradientNormalization.RenormalizeL2PerParamType)
.activation(Activation.RELU)
.learningRate(learningRate)
.weightInit(WeightInit.RELU)
.optimizationAlgo(OptimizationAlgorithm.STOCHASTIC_GRADIENT_DESCENT)
.updater(Updater.RMSPROP)
.list()
.layer(0, new ConvolutionLayer.Builder(new int[] { 5, 5 }, new int[] { 2, 2 }, new int[] { 2, 2 })
.name("inputLayer")
.nIn(channels)
.nOut(96)
.biasInit(0).build())
.layer(1, maxPool("maxpool1"))
.layer(2, new ConvolutionLayer.Builder(new int[] { 5, 5 }, new int[] { 1, 1 }, new int[] { 1, 1 })
.name("convLayer")
.nOut(256)
.biasInit(0).build())
.layer(3, maxPool("maxpool2"))
.layer(4, new ConvolutionLayer.Builder(new int[] { 3, 3 }, new int[] { 1, 1 }, new int[] { 1, 1 })
.name("convLayer2")
.nOut(256)
.biasInit(0).build())
.layer(5, maxPool("maxpool3"))
.layer(6, new DenseLayer.Builder().nOut(512).build())
.layer(7, new DenseLayer.Builder().nOut(256).build())
.layer(8, new OutputLayer.Builder(LossFunctions.LossFunction.MCXENT)
.nOut(outputNum)
.activation(Activation.SOFTMAX).build())
.setInputType(InputType.convolutional(height, width, channels))
.build();
MultiLayerNetwork model = new MultiLayerNetwork(conf);
model.init();
model.setListeners(new ScoreIterationListener(10)); // Print score every 10 parameter updates
EarlyStoppingConfiguration<MultiLayerNetwork> esConf = new EarlyStoppingConfiguration.Builder<MultiLayerNetwork>()
.epochTerminationConditions(new MaxEpochsTerminationCondition(nEpochs))
.evaluateEveryNEpochs(1)
.epochTerminationConditions(new ScoreImprovementEpochTerminationCondition(20))
.scoreCalculator(new DataSetLossCalculator(getDataSetIterator(recordReader, validData, null, batchSize, outputNum), true))
.build();
DataSetIterator dataIter = getDataSetIterator(recordReader, trainData, null, batchSize, outputNum);
EarlyStoppingTrainer trainer = new EarlyStoppingTrainer(esConf, conf, dataIter);
EarlyStoppingResult<MultiLayerNetwork> result = trainer.fit();
ImageTransform flipTransform1 = new FlipImageTransform(rng);
// ImageTransform rotateTransform = new RotateImageTransform(rng, 20, 20, 100, 5);
ImageTransform warpTransform = new WarpImageTransform(rng, 42);
List<ImageTransform> transforms = Arrays.asList(new ImageTransform[] { flipTransform1, warpTransform/* , rotateTransform */ });
for (ImageTransform transform : transforms) {
System.out.print("\nTraining on transformation: " + transform.getClass().toString() + "\n\n");
dataIter = getDataSetIterator(recordReader, trainData, transform, batchSize, outputNum);
trainer = new EarlyStoppingTrainer(esConf, result.getBestModel(), dataIter);
result = trainer.fit();
}
log.info("Evaluate model....");
dataIter = getDataSetIterator(recordReader, testData, null, batchSize, outputNum);
Evaluation eval = model.evaluate(dataIter, labels);
log.info(eval.stats(true));
File modelFile = new File("TrainedModel-" + System.currentTimeMillis() + ".zip");
ModelSerializer.writeModel(model, modelFile, true);
log.info("Model saved to {}.", modelFile);
}
private static DataSetIterator getDataSetIterator(ImageRecordReader recordReader, InputSplit data, ImageTransform transform, int batchSize, int outputNum) {
try {
recordReader.initialize(data, transform);
} catch (IOException e) {
log.warn("Impossible to initialize recordReader", e);
}
DataSetIterator dataIter = new RecordReaderDataSetIterator(recordReader, batchSize, 1, outputNum);
DataNormalization scaler = new ImagePreProcessingScaler(-1, 1);
scaler.fit(dataIter);
dataIter.setPreProcessor(scaler);
return dataIter;
}
private static SubsamplingLayer maxPool(String name) {
return new SubsamplingLayer.Builder(new int[] { 2, 2 }, new int[] { 2, 2 }).name(name).build();
}
}