1. Get the actual or true class labels of test images:

test_true_class <- c(unlist(images.lab.test))

2. Get the predicted class labels of test images. Remember to add 1 to each class label, as the starting index of TensorFlow (the same as Python) is 0 and that of R is 1:

test_pred_class <- y_pred_cls$eval(feed_dict = dict(
x = test_data$images, y_true = test_data$labels, keep_prob = 1.0))
test_pred_class <- test_pred_class + 1

3. Generate the confusion matrix with rows as true labels and columns as predicted labels:

table(actual = test_true_class, predicted = test_pred_class)

4. Generate a plot of the confusion matrix:

confusion <- as.data.frame(table(actual = test_true_class, predicted = test_pred_class))
plot <- ggplot(confusion)
plot + geom_tile(aes(x=actual, y=predicted, fill=Freq)) + scale_x_discrete(name="Actual Class") + scale_y_discrete(name="Predicted Class") + scale_fill_gradient(breaks=seq(from=-.5, to=4, by=.2)) + labs(fill="Normalized
Frequency")

5. Run a helper function to plot images:

check.image <- function(images.rgb,index,true_lab, pred_lab) {
require(imager)
# Testing the parsing: Convert each color layer into a matrix,
# combine into an rgb object, and display as a plot
img <- images.rgb[[index]]
img.r.mat <- as.cimg(matrix(img$r, ncol=32, byrow = FALSE))
img.g.mat <- as.cimg(matrix(img$g, ncol=32, byrow = FALSE))
img.b.mat <- as.cimg(matrix(img$b, ncol=32, byrow = FALSE))
img.col.mat <- imappend(list(img.r.mat,img.g.mat,img.b.mat),"c")
# Plot with actual and predicted label
plot(img.col.mat,main=paste0("True: ", true_lab,":: Pred: ",
pred_lab),xaxt="n")
axis(side=1, xaxp=c(10, 50, 4), las=1)
}

6. Plot random misclassified test images:

labels <- c("airplane","automobile","bird","cat","deer","dog","frog","horse","ship","truck")
# Plot misclassified test images
plot.misclass.images <- function(images.rgb, y_actual, y_predicted,labels){
# Get indices of misclassified
indices <- which(!(y_actual == y_predicted))
id <- sample(indices,1)
# plot the image with true and predicted class
true_lab <- labels[y_actual[id]]
pred_lab <- labels[y_predicted[id]]
check.image(images.rgb,index=id, true_lab=true_lab,pred_lab=pred_lab)
}
plot.misclass.images(images.rgb=images.rgb.test,y_actual=test_true_class,y_predicted=test_pred_class,labels=labels)