how to plot roc?

Helo everyone...Iam working on a dataset of 200 images consisting of both pothole and non-pothole images for the pothole detection.Finally my output is a binary segmented image. I have calculated TP,TN.FP,FN at three thresholds respectively.

For example, if the pothole area is covering atleast 40% of black pixels it is a pothole else it is not a pothole. Similarly, the other two thresholds are of 50% and 60%. For this scenario, can any one help to plot roc curve.

Answer 1

[Star Strider]

Perhaps: plotroc?

[Srinidhi Gorityala]

TPR = [0.71333333333 0.79333333333 0.86666666667];

FPR = [0.00000000000 0.00000000000 0.08860759494];

plot(TPR,FPR);

The above is the code i have used for plotting roc curve at 3 different thresholds. The curve looks strange.Did I evaluated the curve correctly?. Could anyone help me in plotting........

[Star Strider]

According to the Wikipedia article on the Receiver operating characteristic (and specifically this plot figure, since it is easier to see) the plot arguments are reversed.

Try this:

TPR = [0.71333333333 0.79333333333 0.86666666667];
FPR = [0.00000000000 0.00000000000 0.08860759494];

figure
plot(FPR, TPR)
grid
axis([0  1    0  1])                                % Optional (Displays Entire Axis Limits)

It would help to have more data to better define the curve.