I want to teach a 2D data set. Is that possible? Background: I have a function curve y=f(x) and would like to have the resulting y (or its derivative) assigned to certain, fixed classes . I have only found examples for 3D, i.e. finding a structure in an image (y=f(x, y)).

letâ€™s rephrase this so the confusing statistics terminology doesnâ€™t stand in the way.

we express stuff as tensors, i.e. scalars (0D), vectors (1D), matrices (2D) and tensors (3D+).

the value dimension is usually not counted. we arenâ€™t considering relations but functions.

your data appears to be a vector. it can be plotted as a curve.

for comparison: 2D data would be grayscale images or depth data or many other things. they can be plotted as grayscale/false-color images, or as 3D plots showing a height field. a color image would be 3D data, with usually three color components in the last/innermost dimension.

your threadâ€™s title says â€śhaar cascadeâ€ť. those are used to locate objects in a picture (2D/grayscale image) by their appearance.

are you asking how to locate patterns in a 1D signal?

detection/localization is different from classification. classification slaps a class label on the signal as a whole.

cascade classifiers are for classification but in OpenCV theyâ€™re only used for detection/localization, using the â€śmoving windowâ€ť approach, on multiple spatial scales, and specifically Haar or LBP features.

OpenCV has some stuff in the `ml` module, which supports training/regression. its `dnn` module is for inference only, not for training.

can you show examples of the data you are dealing with?

1 Like

Thank you first of all for a good and detailed answer.

Yes, I mean vectors then.

As an example, Iâ€™ll just say a series of measurements to be classified. This can be just one graph, but also several that are connected in a mechanical reference system. If, for example, a component no longer works properly, this also has an effect on other components. These do not break immediately, but I would like to use their movement for early detection of a failure.

For example, the vibration that is transmitted to the housing.

I hope that I have expressed myself clearly, as I am using an English translation software.