Index exceeds matrix dimension

I have this error and I do not understand why. Here's the code:

>> I1_inliers

I1_inliers = 

    608×1 SURFPoints array with properties:

                Scale: [1×608 single]
      SignOfLaplacian: [1×608 int8]
          Orientation: [1×608 single]
             Location: [608×2 single]
               Metric: [1×608 single]
                Count: 608

>> I1_inliers(1)

ans = 

    SURFPoints with properties:

                Scale: 2.8000
      SignOfLaplacian: 1
          Orientation: 0.3424
             Location: [753.3676 999.5223]
               Metric: 1.8129e+04
                Count: 1

>> I1_inliers(2)
Index exceeds matrix dimensions.

I have the same error for every index but 1.

[KSSV]

what is length(I1_inliers)??

[John D'Errico]

It looks like someone has written a SURFPoints class, but I might look at how SURFPoints/subsref was implemented.

[Adam]

There's definitely something non-standard in whatever SURFPoints is. When I display an array of custom class objects I just get something like this:

K>> wavelets

wavelets = 

1×13 RickerWaveletHz array with properties:

    frequency
    sampleRate
    phase
    times
    centralFrequency
    amplitudes
    frequencies
    spectrum

which gives no information on the size of arrays within each object (obviously there is no guarantee of fixed size for these arrays across objects). This alone can be changed by giving a custom disp function, but if you are creating custom disp functions then you are likely doing other things as well as John D'Errico suggests.

[Matteo Foglio]

About length

>> length(I1_inliers)
Function 'subsindex' is not defined for values of class 'SURFPoints'.

SURFPoints is a class of the computer vision toolbox. They represent a set of features inside images.

[Adam]

I suspect that it is actually a scalar with an overloaded size function, looking at it again. I have done that myself once or twice on a class (and usually regretted it afterwards because it confuses me too much).

The size function can be overloaded so that in a case like yours where the object contains a lot of arrays all of the same size the 'size' function is programmed to return that size which over-rides the usual size function which would return [1 1] as the size.

So even though it claims to be a 608x1 array it actually isn't and is just a scalar object containing arrays of length 608.

You would need to access the individual properties with indexing in that case.

I don't have the Computer Vision Toolbox though so I am just surmising based on the symptoms you describe and what is shown when you display the object