It seems like we are talking incidental
just another function but this is truly ground breaking. We use a type
of Neural Network to achieve auto tagging. The first reaction I expect
is that anybody who is anybody will say Auto Tagging is impossible.
But we have devised a method over the last four years of research into
picture recognition that will really change the way people think about auto
We have said that currently after a Picture Recognition
Search images are tagged so that they can be easily found in the future.
Because this tagging is done by a human we know that it is 100% correct or
100% correct for the person who did the tagging. We use this knowledge from
previously tagged images to help Auto Tag other images. But what is
impressive about this system is that it is iterative. Once those new
files have been tagged they then can be used to help tag other files and the
process continues until all images are tagged.
The question is asked
how can you get around a 99.9% accuracy levels or a default 97%. (Please note there are
intellectual Property rights applied to this explanation.) When
searching we will only initially take the closest overall match in a folder
- but can be a set of folders or even the whole disk drive, default is
folders. The closest match is then tagged and only if the match is close
enough using picture recognition. On the next iteration the same thing
is done over and over again, therefore the jump is very small indeed
allowing for a high amount of certainty. But the end result is that the software will
be confident enough to auto tag.
However even within auto tagging you
can lower the certainty and allow bigger jumps between images but when doing
this it is advised that the results are manually checked. With Auto
Tagging (NO SAMPLES ARE REQUIRED) it just uses image that have already been
tagged to search out and tag images that have not yet been tagged.
This can be done in the background whilst the user gets on with other
activities. Auto Tagging Demo Video