1
$\begingroup$

Interpolation is a common way to make an image fit the right input shape for a neural network.

But is there any point in using interpolation to make it easier for the network to learn?

I assume interpolation adds no extra information to the input; It only uses existing information to increase the resolution and fill missing values.

However, sometimes I have observed that while I can not see anything with my human eye, using some kind of advanced interpolation technique such as b-spline interpolation makes it crystal clear that the object i am looking for is in the image, especially in the domain of infrared images.

So, is there any benefit for using interpolation rather than feeding a low dimensional image to a neural network?

Improve this question
$\endgroup$

1 Answer 1

Reset to default
1
$\begingroup$

One of the used is to use a neural interpolation, this is using a pretrained model to zoom our images, in some sense training with this kind of images is like combining this pretrainined model with our new data, and can be seen as transfer learning technique

Improve this answer
$\endgroup$
1
  • $\begingroup$ I can't really understand how this answers my question. What are the benefits of using interpolation before feeding the input to the neural network, if there are any, is my question. I am confused reading this answer. Could you explain more clearly? $\endgroup$
    – MattSt
    Commented Mar 9, 2020 at 14:55

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .