Calculating the miss rate for YOLOv2

I used:

[am, fppi, missRate] = evaluateDetectionMissRate(detectionResults, testData);

figure;

loglog(fppi, missRate);

grid on

title(sprintf('Log Average Miss Rate = %.1f', am))

But the Log Average Miss Rate is always = 1

I tried to plot the test images with the predicted bounding boxes, and my model does predict bounding boxes but i don't know why does this function give me the Log Average Miss Rate is always = 1 ???

Answer 1

[Chetan Gupta]

Hi Raghad,

I understand that you are trying to calculate the log miss rate using evaluateDetectionMissRate(). But your second input argument is wrong.

If you are using image data store. You can go for this approach instead

blds = boxLabelDatastore(testData(:,2:end));
[am, fppi, missRate] = evaluateDetectionMissRate(detectionResults,blds);

You can refer to Evaluate miss rate metric for object detection - MATLAB evaluateDetectionMissRate (mathworks.com) for more information about evaluateDetectionMissRate() function and to Datastore for bounding box label data - MATLAB (mathworks.com) for information about boxLabelDatastore.