org.deeplearning4j.zoo.model.VGG16 Maven / Gradle / Ivy
package org.deeplearning4j.zoo.model;
import lombok.AllArgsConstructor;
import lombok.Builder;
import org.deeplearning4j.nn.api.Model;
import org.deeplearning4j.nn.api.OptimizationAlgorithm;
import org.deeplearning4j.nn.conf.CacheMode;
import org.deeplearning4j.nn.conf.ComputationGraphConfiguration;
import org.deeplearning4j.nn.conf.NeuralNetConfiguration;
import org.deeplearning4j.nn.conf.WorkspaceMode;
import org.deeplearning4j.nn.conf.inputs.InputType;
import org.deeplearning4j.nn.conf.layers.ConvolutionLayer;
import org.deeplearning4j.nn.conf.layers.OutputLayer;
import org.deeplearning4j.nn.conf.layers.SubsamplingLayer;
import org.deeplearning4j.nn.graph.ComputationGraph;
import org.deeplearning4j.zoo.ModelMetaData;
import org.deeplearning4j.zoo.PretrainedType;
import org.deeplearning4j.zoo.ZooModel;
import org.deeplearning4j.zoo.ZooType;
import org.nd4j.linalg.activations.Activation;
import org.nd4j.linalg.learning.config.IUpdater;
import org.nd4j.linalg.learning.config.Nesterovs;
import org.nd4j.linalg.lossfunctions.LossFunctions;
/**
* VGG-16, from Very Deep Convolutional Networks for Large-Scale Image Recognition
* https://arxiv.org/abs/1409.1556
*
* Deep Face Recognition
* http://www.robots.ox.ac.uk/~vgg/publications/2015/Parkhi15/parkhi15.pdf
*
* ImageNet weights for this model are available and have been converted from https://github.com/fchollet/keras/tree/1.1.2/keras/applications.
* CIFAR-10 weights for this model are available and have been converted using "approach 2" from https://github.com/rajatvikramsingh/cifar10-vgg16.
* VGGFace weights for this model are available and have been converted from https://github.com/rcmalli/keras-vggface.
*
* @author Justin Long (crockpotveggies)
*/
@AllArgsConstructor
@Builder
public class VGG16 extends ZooModel {
@Builder.Default private long seed = 1234;
@Builder.Default private int[] inputShape = new int[] {3, 224, 224};
@Builder.Default private int numClasses = 0;
@Builder.Default private IUpdater updater = new Nesterovs();
@Builder.Default private CacheMode cacheMode = CacheMode.NONE;
@Builder.Default private WorkspaceMode workspaceMode = WorkspaceMode.ENABLED;
@Builder.Default private ConvolutionLayer.AlgoMode cudnnAlgoMode = ConvolutionLayer.AlgoMode.PREFER_FASTEST;
private VGG16() {}
@Override
public String pretrainedUrl(PretrainedType pretrainedType) {
if (pretrainedType == PretrainedType.IMAGENET)
return "http://blob.deeplearning4j.org/models/vgg16_dl4j_inference.zip";
else if (pretrainedType == PretrainedType.CIFAR10)
return "http://blob.deeplearning4j.org/models/vgg16_dl4j_cifar10_inference.v1.zip";
else if (pretrainedType == PretrainedType.VGGFACE)
return "http://blob.deeplearning4j.org/models/vgg16_dl4j_vggface_inference.v1.zip";
else
return null;
}
@Override
public long pretrainedChecksum(PretrainedType pretrainedType) {
if (pretrainedType == PretrainedType.IMAGENET)
return 3501732770L;
if (pretrainedType == PretrainedType.CIFAR10)
return 2192260131L;
if (pretrainedType == PretrainedType.VGGFACE)
return 2706403553L;
else
return 0L;
}
@Override
public Class modelType() {
return ComputationGraph.class;
}
public ComputationGraphConfiguration conf() {
ComputationGraphConfiguration conf =
new NeuralNetConfiguration.Builder().seed(seed)
.optimizationAlgo(OptimizationAlgorithm.STOCHASTIC_GRADIENT_DESCENT)
.updater(updater)
.activation(Activation.RELU)
.cacheMode(cacheMode)
.trainingWorkspaceMode(workspaceMode)
.inferenceWorkspaceMode(workspaceMode)
.graphBuilder()
.addInputs("in")
// block 1
.layer(0, new ConvolutionLayer.Builder().kernelSize(3, 3).stride(1, 1)
.padding(1, 1).nIn(inputShape[0]).nOut(64)
.cudnnAlgoMode(cudnnAlgoMode).build(), "in")
.layer(1, new ConvolutionLayer.Builder().kernelSize(3, 3).stride(1, 1)
.padding(1, 1).nOut(64).cudnnAlgoMode(cudnnAlgoMode).build(), "0")
.layer(2, new SubsamplingLayer.Builder()
.poolingType(SubsamplingLayer.PoolingType.MAX).kernelSize(2, 2)
.stride(2, 2).build(), "1")
// block 2
.layer(3, new ConvolutionLayer.Builder().kernelSize(3, 3).stride(1, 1)
.padding(1, 1).nOut(128).cudnnAlgoMode(cudnnAlgoMode).build(), "2")
.layer(4, new ConvolutionLayer.Builder().kernelSize(3, 3).stride(1, 1)
.padding(1, 1).nOut(128).cudnnAlgoMode(cudnnAlgoMode).build(), "3")
.layer(5, new SubsamplingLayer.Builder()
.poolingType(SubsamplingLayer.PoolingType.MAX).kernelSize(2, 2)
.stride(2, 2).build(), "4")
// block 3
.layer(6, new ConvolutionLayer.Builder().kernelSize(3, 3).stride(1, 1)
.padding(1, 1).nOut(256).cudnnAlgoMode(cudnnAlgoMode).build(), "5")
.layer(7, new ConvolutionLayer.Builder().kernelSize(3, 3).stride(1, 1)
.padding(1, 1).nOut(256).cudnnAlgoMode(cudnnAlgoMode).build(), "6")
.layer(8, new ConvolutionLayer.Builder().kernelSize(3, 3).stride(1, 1)
.padding(1, 1).nOut(256).cudnnAlgoMode(cudnnAlgoMode).build(), "7")
.layer(9, new SubsamplingLayer.Builder()
.poolingType(SubsamplingLayer.PoolingType.MAX).kernelSize(2, 2)
.stride(2, 2).build(), "8")
// block 4
.layer(10, new ConvolutionLayer.Builder().kernelSize(3, 3).stride(1, 1)
.padding(1, 1).nOut(512).cudnnAlgoMode(cudnnAlgoMode).build(), "9")
.layer(11, new ConvolutionLayer.Builder().kernelSize(3, 3).stride(1, 1)
.padding(1, 1).nOut(512).cudnnAlgoMode(cudnnAlgoMode).build(), "10")
.layer(12, new ConvolutionLayer.Builder().kernelSize(3, 3).stride(1, 1)
.padding(1, 1).nOut(512).cudnnAlgoMode(cudnnAlgoMode).build(), "11")
.layer(13, new SubsamplingLayer.Builder()
.poolingType(SubsamplingLayer.PoolingType.MAX).kernelSize(2, 2)
.stride(2, 2).build(), "12")
// block 5
.layer(14, new ConvolutionLayer.Builder().kernelSize(3, 3).stride(1, 1)
.padding(1, 1).nOut(512).cudnnAlgoMode(cudnnAlgoMode).build(), "13")
.layer(15, new ConvolutionLayer.Builder().kernelSize(3, 3).stride(1, 1)
.padding(1, 1).nOut(512).cudnnAlgoMode(cudnnAlgoMode).build(), "14")
.layer(16, new ConvolutionLayer.Builder().kernelSize(3, 3).stride(1, 1)
.padding(1, 1).nOut(512).cudnnAlgoMode(cudnnAlgoMode).build(), "15")
.layer(17, new SubsamplingLayer.Builder()
.poolingType(SubsamplingLayer.PoolingType.MAX).kernelSize(2, 2)
.stride(2, 2).build(), "16")
// .layer(18, new DenseLayer.Builder().nOut(4096).dropOut(0.5)
// .build())
// .layer(19, new DenseLayer.Builder().nOut(4096).dropOut(0.5)
// .build())
.layer(18, new OutputLayer.Builder(
LossFunctions.LossFunction.NEGATIVELOGLIKELIHOOD).name("output")
.nOut(numClasses).activation(Activation.SOFTMAX) // radial basis function required
.build(), "17")
.setOutputs("18")
.backprop(true).pretrain(false)
.setInputTypes(InputType.convolutionalFlat(inputShape[2], inputShape[1], inputShape[0]))
.build();
return conf;
}
@Override
public ComputationGraph init() {
ComputationGraph network = new ComputationGraph(conf());
network.init();
return network;
}
@Override
public ModelMetaData metaData() {
return new ModelMetaData(new int[][] {inputShape}, 1, ZooType.CNN);
}
@Override
public void setInputShape(int[][] inputShape) {
this.inputShape = inputShape[0];
}
}
© 2015 - 2024 Weber Informatics LLC | Privacy Policy