Suggestions for running face detection on all files in database

[fisketjon]

IMatch recommends that one runs batches of max 20.000 images. What is the best way of keeping track of which images have been "scanned" and which haven't? 

I don't want to do something that triggers a need for metadata writeback of all the image files.

It would be nice if IMatch can split the selected files in batches (maybe letting me decide the batch size) and keep track of the batches until they're all "scanned". 

[Mario]

I don't want to do something that triggers a need for metadata writeback of all the image files.

When a face is found in an image, XMP region info will be created, which makes the file pending.
If a person is assigned, XMP PersonsInImage is updated, which makes the file pending.

If you want to process files in batches, add them to a category. Select the images to process, and when processing is complete, un-assign the files from the category.

Or, create several categories, put 5K files in each, process, delete the category when done.

You can also just use a collection like pin or dot to mark files not processed, and then remove the dot/pin when done.

Many ways to do this without adding features to IMatch.