How to visualize the predicted values of segmentation from softmax output?

I want to know (visualize) what value, network has predicted during training.

i used Dice pixel classfication layer to observe this but instead of predicted value it shows the original image. Its written that T is target which means T is ground truth and Y is predicted which means predicted pixels.

Code:

classdef dicePixelClassificationLayer < nnet.layer.ClassificationLayer

% This layer implements the generalized Dice loss function for training

% semantic segmentation networks.

properties(Constant)

% Small constant to prevent division by zero.

Epsilon = 1e-8;

end

methods

function layer = dicePixelClassificationLayer(name)

% layer = dicePixelClassificationLayer(name) creates a Dice

% pixel classification layer with the specified name.

% Set layer name.

layer.Name = name;

% Set layer description.

layer.Description = 'Dice loss';

end

function loss = forwardLoss(layer, Y, T)

% loss = forwardLoss(layer, Y, T) returns the Dice loss between

% the predictions Y and the training targets T.

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

code used to visualize the predictions Y and the training targets T

T1=T(:,:,1);

Y1=Y(:,:,1);

subplot(1,2,1)

imshow(T1)

subplot(1,2,2)

imshow(Y1)

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

% Weights by inverse of region size.

W = 1 ./ sum(sum(T,1),2).^2;

intersection = sum(sum(Y.*T,1),2);

union = sum(sum(Y.^2 + T.^2, 1),2);

numer = 2*sum(W.*intersection,3) + layer.Epsilon;

denom = sum(W.*union,3) + layer.Epsilon;

% Compute Dice score.

dice = numer./denom;

% Return average Dice loss.

N = size(Y,4);

loss = sum((1-dice))/N;

end

function dLdY = backwardLoss(layer, Y, T)

% dLdY = backwardLoss(layer, Y, T) returns the derivatives of

% the Dice loss with respect to the predictions Y.

% Weights by inverse of region size.

W = 1 ./ sum(sum(T,1),2).^2;

intersection = sum(sum(Y.*T,1),2);

union = sum(sum(Y.^2 + T.^2, 1),2);

numer = 2*sum(W.*intersection,3) + layer.Epsilon;

denom = sum(W.*union,3) + layer.Epsilon;

N = size(Y,4);

dLdY = (2*W.*Y.*numer./denom.^2 - 2*W.*T./denom)./N;

end

end

end

left is groundtruth and right side is origianl image, which supoosed to be predicted one. now how can i see the predicted one?

Answer 1

[Mahesh Taparia]

Hi

The softmax layer gives the probability of the predicted class. To get the segmented result from that, you can put the probabilistic threshold of 0.5. For example:

P=T1(:,:,1)>0.5;

Else, in order to evaluate the segmented result, you can use the 'semanticseg' function. For more information, you can refer this documentation. Hope it wil help!

[Raza Ali]

Thank you for the answer.

What is Y and T?

defined in loss fucntion? Are they ground truth (T) and predicted values (Y)?

[Mahesh Taparia]

Yes, they are ground truth (T) and predicted values (Y).