MATLAB Answers

how to implement groundtruth for medical images using Faster RCNN with DenseNet.

6 views (last 30 days)
Augustin Rajkumar
Augustin Rajkumar on 10 Dec 2019
Hi, I was implement the base paper of https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=8747354. which related of esophageal cancer. while implemented of Faster Rcnn along with groundTruth detection of bounding boxes and predicition of bounding boxes. For IoU, we want both bounding boxes to calculate performance metrics. how can apply this work. everybody can help me.
close all;
clear all;
clc;
%input image
[file,path]=uigetfile('*.jpg','select a input image');
str=strcat(path,file);
I=imread(str);
figure(1),imshow(I);
gray=rgb2gray(I);
figure(2),imshow(gray);
% noise removal
noise=fspecial('gaussian');
f=imfilter(gray,noise);
figure(3);
imshow(f)
%Gabor Filter
gabor=GaborFilterBank(4,4,32,32);
features=GaborFeatures(f,gabor,4,4);
save Features features
%gaborWavelet=mean(features)
%Pre-trained
File='D:\Project_1\dataset'
dataset=fullfile(File,'foldername','dataset')
imgdata=imageDatastore('dataset',...
'IncludeSubFolders',true,...
'LabelSource','foldernames','FileExtensions',{'.jpg','.png', '.tif'});
%count of total img
icount=countEachLabel(imgdata)
minsetCount=min(icount{:,2});
maxImages=60;
mincount=min(maxImages,minsetCount);
%split each label
imds=splitEachLabel(imgdata,mincount,'randomize');
countEachLabel(imds)
%DenseNet
net=densenet201();
figure(6),plot(net);
title('Densenet');
set(gca,'Ylim',[150 170]);
%First Layer
First=net.Layers(1);
%Last Layer
End=net.Layers(end);
%Number of Class names for ImageNet Classification Task
numel(End.ClassNames);
[trainingSet, testSet]=splitEachLabel(imds,0.3,'randomize');
%resize
imageSize=First.InputSize;
augmentedTrainingSet=augmentedImageDatastore(imageSize,trainingSet,'ColorPreprocessing','gray2rgb');
augmentedTestSet=augmentedImageDatastore(imageSize,testSet,'ColorPreprocessing','gray2rgb');
%Layers
layer=[(imageInputLayer([224 224 3]))
convolution2dLayer(5,16,'Padding','Same','Name','conv_1')
batchNormalizationLayer('Name','BN_1')
reluLayer('Name','relu_1')
maxPooling2dLayer(2,'stride',2)
convolution2dLayer(5,16,'Padding','Same','Name','conv_1')
batchNormalizationLayer('Name','BN_1')
reluLayer('Name','relu_1')
maxPooling2dLayer(2,'stride',2)
averagePooling2dLayer(2,'stride',2)
fullyConnectedLayer(4)
softmaxLayer
classificationLayer()];
featureLayer='fc1000';
trainingFeatures=activations(net,augmentedTrainingSet,featureLayer,'MiniBatchSize',32,'OutputAs','columns');
%Get training labels from trainingset
trainingLabels=trainingSet.Labels;
classifier=fitcecoc(trainingFeatures,trainingLabels,'Learners','Linear','Coding','onevsall','ObservationsIn','columns');
%Test Features
testFeatures=activations(net,augmentedTestSet,featureLayer,'MiniBatchSize',32,'OutputAs','columns');
predictedLabels =predict(classifier,testFeatures,'ObservationsIn','columns');
testLabels=testSet.Labels;
% %confusion matrix
% confMat=confusionmat(testLabels,predictedLabels);
% %convert confusionMatrix to precentage
% confMat=bsxfun(@rdivide,confMat,sum(confMat,2))
% mean(diag(confMat))
%
%automatic resize
aids=augmentedImageDatastore(imageSize,f,'ColorPreprocessing','gray2rgb');
% figure(12),imshow(aids)
imageFeatures= activations(net,aids,featureLayer,'OutputAs','columns');
save DensenetFeatures imageFeatures
% % sdf=imageFeatures'
% % save sdf sdf
% predictedLabels=predict(classifier,imageFeatures,'ObservationsIn','columns');
% g=reshape(imageFeatures,100,5,[])
% g=double(g)
% the=resize(g,[256,256])
% datastruct = load('DensenetFeatures.mat');
% fn = fieldnames(datastruct);
% firstvar = fn{1};
% data = datastruct.(firstvar);
% imwrite( data, 'DenseNet.jpg' );
% et=imread('DenseNet.jpg')
% et=double(et)
%
% % z=imresize(et,[256 256])
% z=reshape(et,256,256,[])
% save Dense z
% %
% r=im2double(z)
figure(7),imshow(f);
%concatenated features
load GaborFeatures.mat
load DensenetFeatures.mat
features = cat(3,imageFeatures,Gabor);
save ConcatenatedFeatures features
datastruct = load('ConcatenatedFeatures.mat');
fn1 = fieldnames(datastruct);
firstvar = fn1{1};
data = datastruct.(firstvar);
% imwrite( data, 'concat.jpg' );
% et1=imread('concat.jpg')
% et1=imread(data)
% figure(87),imshow(et1)
% et2=double(et1)
%
% z=imresize(et2,[256 256])
% Faster R-CNN
lgraph = layerGraph(net);
% Remove the last 3 layners.
layersToRemove = {
'fc1000'
'fc1000_softmax'
'ClassificationLayer_fc1000'
};
lgraph = removeLayers(lgraph, layersToRemove);
% Specify the number of classes the network should classify.
numClasses = 2;
numClassesPlusBackground = numClasses + 1;
% Define new classification layers.
newLayers = [
fullyConnectedLayer(numClassesPlusBackground, 'Name', 'rcnnFC')
softmaxLayer('Name', 'rcnnSoftmax')
classificationLayer('Name', 'rcnnClassification')
];
% Add new object classification layers.
lgraph = addLayers(lgraph, newLayers);
% Connect the new layers to the network.
lgraph = connectLayers(lgraph, 'avg_pool', 'rcnnFC');
% Define the number of outputs of the fully connected layer.
numOutputs = 4 * numClasses;
% Create the box regression layers.
boxRegressionLayers = fullyConnectedLayer(numOutputs,'Name','rcnnBoxFC')
% Add the layers to the network.
lgraph = addLayers(lgraph, boxRegressionLayers);
% Connect the regression layers to the layer named 'avg_pool'.
lgraph = connectLayers(lgraph,'avg_pool','rcnnBoxFC');
% Select a feature extraction layer.
% featureExtractionLayer = 'activation_40_relu';
% Disconnect the layers attached to the selected feature extraction layer.
% lgraph = disconnectLayers(lgraph, featureExtractionLayer,'res5a_branch2a');
% Add ROI max pooling layer.
% SZ=size(testImage)
outputSize = [14 14];
% % roiPool = roiMaxPooling2dLayer(outputSize,'Name','roiPool');
% lgraph = addLayers(lgraph, roiPool);
% Connect feature extraction layer to ROI max pooling layer.
% lgraph = connectLayers(lgraph, featureExtractionLayer,'roiPool/in');
% Connect the output of ROI max pool to the disconnected layers from above.
% lgraph = connectLayers(lgraph, 'roiPool','res5a_branch2a');
% lgraph = connectLayers(lgraph, 'roiPool','res5a_branch1');
% Define anchor boxes.
anchorBoxes = [
16 16
32 16
16 32
];
% Create the region proposal layer.
% proposalLayer = regionProposalLayer(anchorBoxes,'Name','regionProposal');
% lgraph = addLayers(lgraph, proposalLayer)
% Number of anchor boxes.
numAnchors = size(anchorBoxes,1);
% Number of feature maps in coming out of the feature extraction layer.
numFilters = 1000;
img= f;
img1 = img < 65;
img2 = imclearborder(img1);
rp = regionprops(img2, 'BoundingBox', 'Area');
area = [rp.Area];
[~,ind] = max(area);
bb = rp(ind).BoundingBox;
figure(8),imshow(img);
%figure(9),imhist(img)
rectangle('Position', bb, 'EdgeColor', 'red');
rpnLayers = [
convolution2dLayer(3, numFilters,'padding',[1 1],'Name','rpnConv3x3')
reluLayer('Name','rpnRelu')
];
lgraph = addLayers(lgraph, rpnLayers);
% Connect to RPN to feature extraction layer.
lgraph = connectLayers(lgraph,'avg_pool','rpnConv3x3');
% Add RPN classification layers.
rpnClsLayers = convolution2dLayer(1, numAnchors*2,'Name', 'rpnConv1x1ClsScores')
lgraph = addLayers(lgraph, rpnClsLayers);
% Connect the classification layers to the RPN network.
lgraph = connectLayers(lgraph, 'rpnRelu', 'rpnConv1x1ClsScores');
% Add RPN regression layers.
rpnRegLayers = convolution2dLayer(1, numAnchors*4, 'Name', 'rpnConv1x1BoxDeltas')
lgraph = addLayers(lgraph, rpnRegLayers);
% Connect the regression layers to the RPN network.
lgraph = connectLayers(lgraph, 'rpnRelu', 'rpnConv1x1BoxDeltas');
% % Connect region proposal network.
% lgraph = connectLayers(lgraph, 'rpnConv1x1ClsScores', 'regionProposal/scores');
% lgraph = connectLayers(lgraph, 'rpnConv1x1BoxDeltas', 'regionProposal/boxDeltas');
%
% % Connect region proposal layer to roi pooling.
% lgraph = connectLayers(lgraph, 'regionProposal', 'roiPool/roi');
%
% figure(6),
% imshow(testImage);
%
% Show the network after adding the RPN layers.
figure(9),
plot(lgraph)
ylim([30 42])
title('network after adding the RPN layers')
%GroundTruth
data=load('gTruth.mat', 'gTruth');
detector = data.gTruth;
% trainingData = gTruth.DataSource;
%
% dataset=fullfile(File,'foldername','dataset')
% trainingData.imageFilename=imageDatastore('dataset',...
% 'IncludeSubFolders',true,...
% 'LabelSource','foldernames','FileExtensions',{'.jpg','.png', '.tif'});
% trainingData.imageFilename = fullfile('dataset','trainingSet',...
% trainingData.imageFilename);
% trainingData.imageFilename = fullfile(toolboxdir('vision'),'visiondata', ...
% trainingData.imageFilename);
% rng(0);
% shuffledIdx = randperm(height(trainingData));
% trainingData = trainingData(shuffledIdx,:);
%
% imds = imageDatastore(trainingData.imageFilename);
% % Create a box label datastore using the label columns from the table.
%
% blds = boxLabelDatastore(trainingData(:,2:end));
% % Combine the datastores.
%
% ds = combine(imds, blds);
% % Set up the network layers.
%
% lgraph = layerGraph(data.detector.Network)
% Configure training options.
% options = trainingOptions('sgdm', ...
% 'MiniBatchSize', 1, ...
% 'InitialLearnRate', 1e-3, ...
% 'MaxEpochs', 7, ...
% 'VerboseFrequency', 200, ...
% 'CheckpointPath', tempdir);
%
% detector = trainFasterRCNNObjectDetector(gTruth,ppp, options, ...
% 'NegativeOverlapRange',[0 0.3], ...
% 'PositiveOverlapRange',[0.6 1], ...
% 'SmallestImageDimension',300);
%
[bbox, score, label] = detect(detector,img);
detectedImg = insertShape(img,'Rectangle',bbox);
figure(11)
imshow(detectedImg)

  0 Comments

Sign in to comment.

Answers (0)

Sign in to answer this question.