Error using ./ in train faster RCNN

Hi everyone,

i try to train a faster rcnn detector, but i got some error like this:

Error using ./

Matrix dimensions must agree.

Error in vision.internal.cnn.layer.RPNCrossEntropy/backwardLoss (line 161)

dX = (-T./nnet.internal.cnn.util.boundAwayFromZero(Y))./numObservations;

Error in vision.internal.cnn.layer.RPNOutputInternalLayer/backwardLoss (line 99)

dLdZCls = this.RPNClassificationLayer.backwardLoss(Zcls, Tcls);

Error in nnet.internal.cnn.DAGNetwork/computeGradientsForTraining/efficientBackProp (line 748)

dLossdX = thisLayer.backwardLoss( ...

Error in nnet.internal.cnn.DAGNetwork>@()efficientBackProp(i) (line 840)

@() efficientBackProp(i), ...

Error in nnet.internal.cnn.util.executeWithStagedGPUOOMRecovery (line 11)

[ varargout{1:nOutputs} ] = computeFun();

Error in nnet.internal.cnn.DAGNetwork>iExecuteWithStagedGPUOOMRecovery (line 1563)

[varargout{1:nargout}] = nnet.internal.cnn.util.executeWithStagedGPUOOMRecovery(varargin{:});

Error in nnet.internal.cnn.DAGNetwork/computeGradientsForTraining (line 839)

theseGradients = iExecuteWithStagedGPUOOMRecovery( ...

Error in nnet.internal.cnn.Trainer/computeGradients (line 200)

[gradients, predictions, states] = net.computeGradientsForTraining(X, Y, propagateState);

Error in nnet.internal.cnn.Trainer/train (line 119)

[gradients, predictions, states] = this.computeGradients(net, X, response, propagateState);

Error in vision.internal.cnn.trainNetwork (line 96)

trainedNet = trainer.train(trainedNet, trainingDispatcher);

Error in trainFasterRCNNObjectDetector>iTrainEndToEnd (line 897)

[net, info] = vision.internal.cnn.trainNetwork(...

Error in trainFasterRCNNObjectDetector (line 428)

[detector, info] = iTrainEndToEnd(trainingData, fastRCNN, options, params, executionSettings, imageInfo);

Error in TrainDetectorFasterRCNNv12 (line 94)

[rcnn,info] = trainFasterRCNNObjectDetector(preprocessedTrainingData, lgraph, opts, ...

i know some error about './' in basic operational matrix. but i don't have idea where i supose to be fixed in this case.

Is there a solution to this problem?

Thanks

[Birju Patel]

@Septian NLP - are you still running in to this issue? If so, can you upload your ground truth data and any other files required to run your code? That will help me reproduce the issue to see if I can provide a fix.

Answer 1

[yanqi liu]

yes，sir，may be make image size to same，such as augmentedImageDatastore make image to 227*227