In computer vision and image processing, we say a method to recognise an object in an image is position independent if it works no matter where the target is within the image. This may be achieved by training on lots of different offsets of the same images, by centering images on pertinent features, or by using models that are intrinsically position independent. In additoon we often seek scale independence and rotational independence (it doesnt matter how big or what anfgle the object is).
Used on page 277