Showing pictures that have only one specific keyword, e.g. a specific person

Started by blackhead2, October 22, 2017, 11:47:20 AM

Previous topic - Next topic

blackhead2

Hello together,

at the moment I'm using the trial version of IMatch 2017 and it looks very promising but I have a problem I couldn't figure out by myself.

I try to create a category that only shows pictures of one specific person. Pictures with two or more persons should not be shown.

With the example from the help I can achieve this by specifying the person and excluding the others, e.g.:

|-- Person
   |-- Lisa
   |-- John
   ...

With the category formula "Person|Lisa" NOT "Person|John" I get the correct results but I have to adapt the formula everytime I add a new person which is not what I want.

I tried the follwoing formula which is not working:

"Person|Lisa" NOT ( "Person" NOT "Person|Lisa"). The part in the brackets shows all pictures without "Lisa" so I thought that might work but it doesn't.

I there a way to achieve a generic solution for this question? Or is there maybe just a logical error in my fomula?!

Thank you in advance!

Best regards

Jens


Jingo

Welcome aboard... the power of IMatch (and its community / Developer) is unparalleled.... hope you decide to purchase and join the fun!

Mario

Another solution for this I have seen (depends on how you work in general) is to assign additional info to a file, e.g. "single", "couple", "group". This not only allows you to easily find photos of one person, two persons or groups, but also allows you to find images showing only "Lisa", by combining it with the "single" category.

I expect classifications like "one person", "two persons", ... to be provided automatically once I include the support for 3rd party A.I. modules. Several of the provides return the number of persons detected in an image as a result, and this can then be used by IMatch for many purposes, including the one outlined above  :)
-- Mario
IMatch Developer
Forum Administrator
http://www.photools.com  -  Contact & Support - Follow me on 𝕏 - Like photools.com on Facebook

blackhead2

Thank you for your replies. As I now know how to categorize pictures with a single person it is at least easy to add the additional info ;) Till now I prefer tagging the persons manually not using any AI as it is (at least in my opinion) difficult to find mistakes made by the AI. Photos where you can only see the back of a person e.g. in some landscape photos are often not properly detected. Especially if only one person in a group is not detected. And checking every picture after using the AI is not what I want. But this is only my experience with face detection using picasa and windows photo gallery. Maybe there are far better solutions.

Mario

QuoteAnd checking every picture after using the AI is not what I want. But this is only my experience with face detection using picasa and windows photo gallery. Maybe there are far better solutions.

That's a general problem with all A.I. based algorithms. You always have to check manually afterwards if there are no false positives, or plain nonsense.
The big vendors tell you that in their shiny adds ("upload all your files into the A... cloud and we will tag your files automatically"), though.

When I implement this for IMatch or IMatch Anywhere I plan to have an interactive mode where you go from image to image, IMatch presents you the results of the A.I. and you can acknowledge them, reject them or untag individual keyword suggestions. This is basically A.I. supported tagging and captioning. There will also be a batch mode, if the chosen vendor offers that.

The A.I. results are sometimes truly amazing and sometimes so-la-la and sometimes plain wrong. Still, it can save you tons of time if you just have to okay results instead of typing/picking from the thesaurus. And they get better all the time, while the costs get low enough so even normal people can afford to use them (1 dollar per 1000 files or so).
-- Mario
IMatch Developer
Forum Administrator
http://www.photools.com  -  Contact & Support - Follow me on 𝕏 - Like photools.com on Facebook