org.deeplearning4j.nn.layers.convolution.ZeroPaddingLayer 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.nn.layers.convolution;
import lombok.val;
import org.deeplearning4j.nn.api.Layer;
import org.deeplearning4j.nn.conf.NeuralNetConfiguration;
import org.deeplearning4j.nn.gradient.DefaultGradient;
import org.deeplearning4j.nn.gradient.Gradient;
import org.deeplearning4j.nn.layers.AbstractLayer;
import org.nd4j.linalg.api.buffer.DataType;
import org.nd4j.linalg.api.ndarray.INDArray;
import org.nd4j.linalg.indexing.INDArrayIndex;
import org.nd4j.linalg.indexing.NDArrayIndex;
import org.nd4j.linalg.primitives.Pair;
import org.deeplearning4j.nn.workspace.LayerWorkspaceMgr;
import org.deeplearning4j.nn.workspace.ArrayType;
/**
* Zero padding layer for convolutional neural networks.
* Allows padding to be done separately for top/bottom/left/right
*
* @author Alex Black
*/
public class ZeroPaddingLayer extends AbstractLayer {
private int[] padding; //[padTop, padBottom, padLeft, padRight]
public ZeroPaddingLayer(NeuralNetConfiguration conf, DataType dataType) {
super(conf, dataType);
this.padding = ((org.deeplearning4j.nn.conf.layers.ZeroPaddingLayer) conf.getLayer()).getPadding();
}
@Override
public boolean isPretrainLayer() {
return false;
}
@Override
public void clearNoiseWeightParams() {
//No op
}
@Override
public Type type() {
return Type.CONVOLUTIONAL;
}
@Override
public Pair backpropGradient(INDArray epsilon, LayerWorkspaceMgr workspaceMgr) {
assertInputSet(true);
val inShape = input.shape();
INDArray epsNext = epsilon.get(NDArrayIndex.all(), NDArrayIndex.all(),
NDArrayIndex.interval(padding[0], padding[0] + inShape[2]),
NDArrayIndex.interval(padding[2], padding[2] + inShape[3]));
epsNext = workspaceMgr.leverageTo(ArrayType.ACTIVATION_GRAD, epsNext);
return new Pair<>((Gradient) new DefaultGradient(), epsNext);
}
@Override
public INDArray activate(boolean training, LayerWorkspaceMgr workspaceMgr) {
assertInputSet(false);
val inShape = input.shape();
val outH = inShape[2] + padding[0] + padding[1];
val outW = inShape[3] + padding[2] + padding[3];
val outShape = new long[] {inShape[0], inShape[1], outH, outW};
INDArray out = workspaceMgr.create(ArrayType.ACTIVATIONS, input.dataType(), outShape, 'c');
out.put(new INDArrayIndex[] {NDArrayIndex.all(), NDArrayIndex.all(),
NDArrayIndex.interval(padding[0], padding[0] + inShape[2]),
NDArrayIndex.interval(padding[2], padding[2] + inShape[3])}, input);
return out;
}
@Override
public Layer clone() {
return new ZeroPaddingLayer(conf.clone(), dataType);
}
@Override
public double calcRegularizationScore(boolean backpropParamsOnly){
return 0;
}
}