[OFFICIAL] IMatch 2020 Discussion Thread

[jmsantos]

Hello Mario.

I wrote a post about Metadata conciliation in May 2019:
https://www.photools.com/community/index.php?topic=9039.0

After several answers, you said:

What we would need to handle your use case is a feature to a) Produce a rich XMP record from the existing data in the file, apply all MWG mappings to map EXIF/GPS/IPTC from the file into the XMP record. And then, if an XMP sidecar file exists, merge its contents selectively with the rich XMP data produced by IMatch. Such a feature currently does not exist.
I have some ideas in that direction for IMatch 2020, though. Maybe even earlier.

This requires a lot of testing, because there are many fringe cases and metadata handling has become more and more complex over time, many options have been added to handle fringe cases and user-specific workflows. Camera vendors have started to add partial XMP records to RAW files etc. I need to give this a re-think for IMatch 2020.

Has Imatch 2020 improved or resolved this issue?

In advance, thanks you for your great work.

[Mario]

Nothing in that area.
Does the standard metadata processing in IMatch not work for you? It seems to work pretty well for users...

[jmsantos]

Well, maybe a institutional Photographic Archive is not a standard user. I thought it was clear in that discussion thread.

I just hoped some progress in those "ideas in that direction for IMatch 2020". Perhaps there is an opportunity for improvement here.

Thanks

[Mario]

Can you run the Metadata Analyst app on one of your files so we see what they contain?
When I understand the old thread correctly your files have a mix of XMP and native metadata, in wrong places and out-of-sync?

[kiwilink]

Do you sign up for "Google Cloud Platform" to use the IMATCH Autotagger?  I saw the different ones and thought the Google one was better because it recognizes Lat/Long but its not clear to me what to sign up for when I am directed to their site..

Thanks!

Kiwilink

[Mario]

This question for me?

Google Vision API is what you need.
Google makes it quite hard to get things working because their products are not aimed at end users. And, even worse, they change the process and look and feel and UI all the time. Nerds.

Log into your Google account https://accounts.google.com/ and enable the API.
See https://cloud.google.com/vision and click on "Get Started fir Free".

Note that Capterra and imagga are more user-frienly. Google does not care for individual users, they want only businesses.
But that would mean that I would have to make all this work and then charge you every month for your AutoTagger usage. I won't go there. Legal and support nightmare.

[Mees Dekker]

IMatch 2020 is running fine on my computers (desktop and travelling laptop), although I will to dive deeper to understand all the new features. I already had "my own system" for finding people: a very extensive category setup (like: persons|family|name). That may help me to setup the people view. Same for events: these were also "categorized".

These new features may be a great help. Although the old system worked for me, it was also rather time-consuming to assign every picture to the wright categories.

Besides these new features, it seems to me that IMatch 2020 is running faster/smoother than former versions/releases. I did not do any explicit tests on this (and cannot do that, since you can't run two versions of IMatch on the same computer). What is the impression of others?

[kiwilink]

Mario:

Thanks for the reply about the Google Vision API.  Signed up and tried the Auto Tagger interface with IMATCH and it did a great job with the keywords.  I was more interested in seeing if the API would recognize landmarks and assign a latitude/longitude to my image.  I selected several well known images shot on the strip in Las Vegas but it did not return any coordinates.  Is Auto Tagger the best choice for me if I'm looking for something to identify and assign lat/long to my pics?  For example, images of the Sydney Opera House, Castles in Germany, Venice, etc? 

Thank You!

Kiwilink

[Mario]

Have you enabled Landmarks for the Google service in AutoTagger options?
This is optional and Google will charge twice per image (once for the keywords, once for the landmarks).

I've just did a few tests with known landmarks in Berlin, London, New York and Google ways always able to pin a landmark with GPS coordinates to the image.
These are the settings I use:

[kiwilink]

Maro:

Thanks for the reply.  Yes, that is the same settings I am using.  I will check a few more.  Did it return a Latitude and longitude on yours.

Thanks

Kiwilink

[Mario]

Yes.

If you run the App in your Web Browser, you can see the responses from Google in the console after opening the developer tools in the browser with F12.
I don't know for how many known landmarks Google has coordinates in their database, though.

If you want, you can send me one of your images (small size is enough) to support email address or attach it here. I can then have a closer look.

[kiwilink]

Mario:

Thanks!  I used this Image.  It had no Keywords and no Latitude/Longitude.  I used IMATCH Autotagger and submitted it with keywords, Landmarks, and Coordinates checked.  It knew it was Chicago (Navy Pier) by the keywords but it did not return any geo latitudes or longitudes.  I tried more famous landmarks like the Sydney Bridge and it did find the Geo Coordinates. 

I also noticed that if I want just the Geo coordinates and I only check Landmarks and apply GPS coordinates I get the attached snap error.  Does that mean I cannot just ask for Landmark and GPS coordinates (and not ask for keywords)?

Thanks!

Kiwilink

[jmsantos]

Can you run the Metadata Analyst app on one of your files so we see what they contain?
When I understand the old thread correctly your files have a mix of XMP and native metadata, in wrong places and out-of-sync?

Yes, I have run Metadata Analyst, although it reports something I already know.
I think it would be necessary to implement a tool to combine the metadata embedded in the file (such as GPS data) and the data in the XMP file. Or already exists in Imatch? I don't know.

[Mario]

combine the metadata embedded in the file (such as GPS data)

IMatch imports native GPS data into XMP during ingest. So the GPS data you see in IMatch is what's in the file. Unless you have modified the GPS data in IMatch but not written back.

[Mario]

Mario:

Thanks!  I used this Image.  It had no Keywords and no Latitude/Longitude.  I used IMATCH Autotagger and submitted it with keywords, Landmarks, and Coordinates checked.  It knew it was Chicago (Navy Pier) by the keywords but it did not return any geo latitudes or longitudes.  I tried more famous landmarks like the Sydney Bridge and it did find the Geo Coordinates. 

I also noticed that if I want just the Geo coordinates and I only check Landmarks and apply GPS coordinates I get the attached snap error.  Does that mean I cannot just ask for Landmark and GPS coordinates (and not ask for keywords)?

Thanks!

Kiwilink

This should become bug report, then?

[Mario]

Mario:

Thanks!  I used this Image.  It had no Keywords and no Latitude/Longitude.  I used IMATCH Autotagger and submitted it with keywords, Landmarks, and Coordinates checked.  It knew it was Chicago (Navy Pier) by the keywords but it did not return any geo latitudes or longitudes.  I tried more famous landmarks like the Sydney Bridge and it did find the Geo Coordinates. 
Thanks!

Kiwilink

Google does not return GPS coordinates for the sample image you have attached. Probable because it is only a crop of the building.

I shall think of adding this info to the auto tagger so the user knows. Maybe a "NO GPS data found" or something below the result list.

[kiwilink]

Mario:

Thank you for checking this!

Kiwilink

[Mario]

This thread has been locked.

Please open new threads if you want to discuss something related to IMatch 2020.