org.deeplearning4j.zoo.model.InceptionResNetV1 Maven / Gradle / Ivy
/*******************************************************************************
* Copyright (c) 2015-2018 Skymind, Inc.
*
* This program and the accompanying materials are made available under the
* terms of the Apache License, Version 2.0 which is available at
* https://www.apache.org/licenses/LICENSE-2.0.
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the
* License for the specific language governing permissions and limitations
* under the License.
*
* SPDX-License-Identifier: Apache-2.0
******************************************************************************/
package org.deeplearning4j.zoo.model;
import lombok.AllArgsConstructor;
import lombok.Builder;
import lombok.NoArgsConstructor;
import org.deeplearning4j.nn.api.Model;
import org.deeplearning4j.nn.api.OptimizationAlgorithm;
import org.deeplearning4j.nn.conf.*;
import org.deeplearning4j.nn.conf.distribution.NormalDistribution;
import org.deeplearning4j.nn.conf.distribution.TruncatedNormalDistribution;
import org.deeplearning4j.nn.conf.graph.L2NormalizeVertex;
import org.deeplearning4j.nn.conf.graph.MergeVertex;
import org.deeplearning4j.nn.conf.inputs.InputType;
import org.deeplearning4j.nn.conf.layers.*;
import org.deeplearning4j.nn.graph.ComputationGraph;
import org.deeplearning4j.nn.weights.WeightInit;
import org.deeplearning4j.zoo.ModelMetaData;
import org.deeplearning4j.zoo.PretrainedType;
import org.deeplearning4j.zoo.ZooModel;
import org.deeplearning4j.zoo.ZooType;
import org.deeplearning4j.zoo.model.helper.InceptionResNetHelper;
import org.nd4j.linalg.activations.Activation;
import org.nd4j.linalg.learning.config.IUpdater;
import org.nd4j.linalg.learning.config.RmsProp;
import org.nd4j.linalg.lossfunctions.LossFunctions;
/**
* A variant of the original FaceNet model that relies on embeddings and triplet loss.
* Reference: https://arxiv.org/abs/1503.03832
* Also based on the OpenFace implementation:
* http://reports-archive.adm.cs.cmu.edu/anon/2016/CMU-CS-16-118.pdf
*
* Revised and consolidated version by @crockpotveggies
*/
@AllArgsConstructor
@Builder
public class InceptionResNetV1 extends ZooModel {
@Builder.Default private long seed = 1234;
@Builder.Default private int[] inputShape = new int[] {3, 160, 160};
@Builder.Default private int numClasses = 0;
@Builder.Default private IUpdater updater = new RmsProp(0.1, 0.96, 0.001);
@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 InceptionResNetV1() {}
@Override
public String pretrainedUrl(PretrainedType pretrainedType) {
return null;
}
@Override
public long pretrainedChecksum(PretrainedType pretrainedType) {
return 0L;
}
@Override
public Class modelType() {
return ComputationGraph.class;
}
@Override
public ComputationGraph init() {
int embeddingSize = 128;
ComputationGraphConfiguration.GraphBuilder graph = graphBuilder("input1");
graph.addInputs("input1").setInputTypes(InputType.convolutional(inputShape[2], inputShape[1], inputShape[0]))
// Logits
.addLayer("bottleneck", new DenseLayer.Builder().nIn(5376).nOut(embeddingSize).build(),
"avgpool")
// Embeddings
.addVertex("embeddings", new L2NormalizeVertex(new int[] {1}, 1e-10), "bottleneck")
// Output
.addLayer("outputLayer",
new CenterLossOutputLayer.Builder()
.lossFunction(LossFunctions.LossFunction.NEGATIVELOGLIKELIHOOD)
.activation(Activation.SOFTMAX).alpha(0.9).lambda(1e-4)
.nIn(embeddingSize).nOut(numClasses).build(),
"embeddings")
.setOutputs("outputLayer");
ComputationGraphConfiguration conf = graph.build();
ComputationGraph model = new ComputationGraph(conf);
model.init();
return model;
}
public ComputationGraphConfiguration.GraphBuilder graphBuilder(String input) {
ComputationGraphConfiguration.GraphBuilder graph = new NeuralNetConfiguration.Builder().seed(seed)
.activation(Activation.RELU)
.optimizationAlgo(OptimizationAlgorithm.STOCHASTIC_GRADIENT_DESCENT)
.updater(updater)
.weightInit(new TruncatedNormalDistribution(0.0, 0.5))
.l2(5e-5)
.miniBatch(true)
.cacheMode(cacheMode)
.trainingWorkspaceMode(workspaceMode)
.inferenceWorkspaceMode(workspaceMode)
.convolutionMode(ConvolutionMode.Truncate).graphBuilder();
graph
// stem
.addLayer("stem-cnn1",
new ConvolutionLayer.Builder(new int[] {3, 3}, new int[] {2, 2})
.nIn(inputShape[0]).nOut(32)
.cudnnAlgoMode(cudnnAlgoMode).build(),
input)
.addLayer("stem-batch1",
new BatchNormalization.Builder(false).decay(0.995).eps(0.001).nIn(32).nOut(32)
.build(),
"stem-cnn1")
.addLayer("stem-cnn2",
new ConvolutionLayer.Builder(new int[] {3, 3}).nIn(32).nOut(32)
.cudnnAlgoMode(cudnnAlgoMode).build(),
"stem-batch1")
.addLayer("stem-batch2",
new BatchNormalization.Builder(false).decay(0.995).eps(0.001).nIn(32).nOut(32)
.build(),
"stem-cnn2")
.addLayer("stem-cnn3",
new ConvolutionLayer.Builder(new int[] {3, 3})
.convolutionMode(ConvolutionMode.Same).nIn(32).nOut(64)
.cudnnAlgoMode(cudnnAlgoMode).build(),
"stem-batch2")
.addLayer("stem-batch3", new BatchNormalization.Builder(false).decay(0.995).eps(0.001).nIn(64)
.nOut(64).build(), "stem-cnn3")
.addLayer("stem-pool4", new SubsamplingLayer.Builder(SubsamplingLayer.PoolingType.MAX,
new int[] {3, 3}, new int[] {2, 2}).build(), "stem-batch3")
.addLayer("stem-cnn5",
new ConvolutionLayer.Builder(new int[] {1, 1}).nIn(64).nOut(80)
.cudnnAlgoMode(cudnnAlgoMode).build(),
"stem-pool4")
.addLayer("stem-batch5",
new BatchNormalization.Builder(false).decay(0.995).eps(0.001).nIn(80).nOut(80)
.build(),
"stem-cnn5")
.addLayer("stem-cnn6",
new ConvolutionLayer.Builder(new int[] {3, 3}).nIn(80).nOut(128)
.cudnnAlgoMode(cudnnAlgoMode).build(),
"stem-batch5")
.addLayer("stem-batch6",
new BatchNormalization.Builder(false).decay(0.995).eps(0.001).nIn(128).nOut(128)
.build(),
"stem-cnn6")
.addLayer("stem-cnn7",
new ConvolutionLayer.Builder(new int[] {3, 3}, new int[] {2, 2}).nIn(128)
.nOut(192).cudnnAlgoMode(cudnnAlgoMode)
.build(),
"stem-batch6")
.addLayer("stem-batch7", new BatchNormalization.Builder(false).decay(0.995).eps(0.001).nIn(192)
.nOut(192).build(), "stem-cnn7");
// 5xInception-resnet-A
InceptionResNetHelper.inceptionV1ResA(graph, "resnetA", 5, 0.17, "stem-batch7");
// Reduction-A
graph
// 3x3
.addLayer("reduceA-cnn1",
new ConvolutionLayer.Builder(new int[] {3, 3}, new int[] {2, 2}).nIn(192)
.nOut(192).cudnnAlgoMode(cudnnAlgoMode)
.build(),
"resnetA")
.addLayer("reduceA-batch1",
new BatchNormalization.Builder(false).decay(0.995).eps(0.001).nIn(192).nOut(192)
.build(),
"reduceA-cnn1")
// 1x1 -> 3x3 -> 3x3
.addLayer("reduceA-cnn2",
new ConvolutionLayer.Builder(new int[] {1, 1})
.convolutionMode(ConvolutionMode.Same).nIn(192).nOut(128)
.cudnnAlgoMode(cudnnAlgoMode).build(),
"resnetA")
.addLayer("reduceA-batch2",
new BatchNormalization.Builder(false).decay(0.995).eps(0.001).nIn(128).nOut(128)
.build(),
"reduceA-cnn2")
.addLayer("reduceA-cnn3",
new ConvolutionLayer.Builder(new int[] {3, 3})
.convolutionMode(ConvolutionMode.Same).nIn(128).nOut(128)
.cudnnAlgoMode(cudnnAlgoMode).build(),
"reduceA-batch2")
.addLayer("reduceA-batch3",
new BatchNormalization.Builder(false).decay(0.995).eps(0.001).nIn(128).nOut(128)
.build(),
"reduceA-cnn3")
.addLayer("reduceA-cnn4",
new ConvolutionLayer.Builder(new int[] {3, 3}, new int[] {2, 2}).nIn(128)
.nOut(192).cudnnAlgoMode(cudnnAlgoMode)
.build(),
"reduceA-batch3")
.addLayer("reduceA-batch4",
new BatchNormalization.Builder(false).decay(0.995).eps(0.001).nIn(192).nOut(192)
.build(),
"reduceA-cnn4")
// maxpool
.addLayer("reduceA-pool5",
new SubsamplingLayer.Builder(SubsamplingLayer.PoolingType.MAX, new int[] {3, 3},
new int[] {2, 2}).build(),
"resnetA")
// -->
.addVertex("reduceA", new MergeVertex(), "reduceA-batch1", "reduceA-batch4", "reduceA-pool5");
// 10xInception-resnet-B
InceptionResNetHelper.inceptionV1ResB(graph, "resnetB", 10, 0.10, "reduceA");
// Reduction-B
graph
// 3x3 pool
.addLayer("reduceB-pool1",
new SubsamplingLayer.Builder(SubsamplingLayer.PoolingType.MAX, new int[] {3, 3},
new int[] {2, 2}).build(),
"resnetB")
// 1x1 -> 3x3
.addLayer("reduceB-cnn2",
new ConvolutionLayer.Builder(new int[] {1, 1})
.convolutionMode(ConvolutionMode.Same).nIn(576).nOut(256)
.cudnnAlgoMode(cudnnAlgoMode).build(),
"resnetB")
.addLayer("reduceB-batch1",
new BatchNormalization.Builder(false).decay(0.995).eps(0.001).nIn(256).nOut(256)
.build(),
"reduceB-cnn2")
.addLayer("reduceB-cnn3",
new ConvolutionLayer.Builder(new int[] {3, 3}, new int[] {2, 2}).nIn(256)
.nOut(256).cudnnAlgoMode(cudnnAlgoMode)
.build(),
"reduceB-batch1")
.addLayer("reduceB-batch2",
new BatchNormalization.Builder(false).decay(0.995).eps(0.001).nIn(256).nOut(256)
.build(),
"reduceB-cnn3")
// 1x1 -> 3x3
.addLayer("reduceB-cnn4",
new ConvolutionLayer.Builder(new int[] {1, 1})
.convolutionMode(ConvolutionMode.Same).nIn(576).nOut(256)
.cudnnAlgoMode(cudnnAlgoMode).build(),
"resnetB")
.addLayer("reduceB-batch3",
new BatchNormalization.Builder(false).decay(0.995).eps(0.001).nIn(256).nOut(256)
.build(),
"reduceB-cnn4")
.addLayer("reduceB-cnn5",
new ConvolutionLayer.Builder(new int[] {3, 3}, new int[] {2, 2}).nIn(256)
.nOut(256).cudnnAlgoMode(cudnnAlgoMode)
.build(),
"reduceB-batch3")
.addLayer("reduceB-batch4",
new BatchNormalization.Builder(false).decay(0.995).eps(0.001).nIn(256).nOut(256)
.build(),
"reduceB-cnn5")
// 1x1 -> 3x3 -> 3x3
.addLayer("reduceB-cnn6",
new ConvolutionLayer.Builder(new int[] {1, 1})
.convolutionMode(ConvolutionMode.Same).nIn(576).nOut(256)
.cudnnAlgoMode(cudnnAlgoMode).build(),
"resnetB")
.addLayer("reduceB-batch5",
new BatchNormalization.Builder(false).decay(0.995).eps(0.001).nIn(256).nOut(256)
.build(),
"reduceB-cnn6")
.addLayer("reduceB-cnn7",
new ConvolutionLayer.Builder(new int[] {3, 3})
.convolutionMode(ConvolutionMode.Same).nIn(256).nOut(256)
.cudnnAlgoMode(cudnnAlgoMode).build(),
"reduceB-batch5")
.addLayer("reduceB-batch6",
new BatchNormalization.Builder(false).decay(0.995).eps(0.001).nIn(256).nOut(256)
.build(),
"reduceB-cnn7")
.addLayer("reduceB-cnn8",
new ConvolutionLayer.Builder(new int[] {3, 3}, new int[] {2, 2}).nIn(256)
.nOut(256).cudnnAlgoMode(cudnnAlgoMode)
.build(),
"reduceB-batch6")
.addLayer("reduceB-batch7",
new BatchNormalization.Builder(false).decay(0.995).eps(0.001).nIn(256).nOut(256)
.build(),
"reduceB-cnn8")
// -->
.addVertex("reduceB", new MergeVertex(), "reduceB-pool1", "reduceB-batch2", "reduceB-batch4",
"reduceB-batch7");
// 10xInception-resnet-C
InceptionResNetHelper.inceptionV1ResC(graph, "resnetC", 5, 0.20, "reduceB");
// Average pooling
graph.addLayer("avgpool",
new SubsamplingLayer.Builder(SubsamplingLayer.PoolingType.AVG, new int[] {1, 1}).build(),
"resnetC");
return graph;
}
@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