Comparing a Tensor to a group of others

1

Hey!

I am quite new here and not really sure if Pytorch is something I am looking for but I found the PyTorch Tensor compare function and it goes into the right direction.

My problem is the following. I do have several (physical) fingerprints in tensor format.
These Tensors have a size of 192x192 values.

I have “known Fingerprints” where I know the owner from, lets say 2 different Fingers a 3 Images. These three Images are similar but differ in the sense that the finger might have been rotated or shifted on the sensor.

I would know need a function that compares the “unknown” fingerprint to the known Fingerprints and rotates and shifts the Tensor to see if it fits and output a confidence factor.

I am programming in Python. Could anyone point me in the right direction?

Thanks!

2

It sounds like what you want is actually an image processing library, not necessarily a ML library, which is what PyTorch is. So you might be able to do what you want using PyTorch, but something like OpenCV might be better suited for it.

Based on your description, it sounds like what you want is a ‘rotation invariant template matching’. OpenCV should have support for that and you should also be able to find a lot of resources online to do this as well.