Face Arrangement

What it Does

IMatch analyzes the number, size and arrangement of faces in your images. This tells IMatch if an image shows one, two, multiple or groups of persons. The analysis also reveals details like if the image is most likely a portrait or a close-up or if the faces are close together. The result of this analysis is accessible via the {File.Faces.Arrangement.*} variables and in other features.

IMatch tries to find the most relevant type of grouping for files. It favors groups over individual faces and it labels two faces within group distance as Two Faces instead of Multiple Faces. If an image shows multiple faces but some of the faces form a group, the file is labeled as Group of Faces. If there is both a group of faces and a large group of faces, the file is labeled as Large Group of Faces.

The following collection of images shows some examples and the result of the analysis:

Keep in mind that the automatic analysis is not perfect. There will always be images that produce a less than optimal result. This is normal and expected. The general goal is that this analysis enables you to organize your files automatically, without manual work. And to use the result an additional way of grouping related files together, for filtering, automatic descriptions etc.

The {File.Faces.Arrangement.*} Variables

These variables return the results of the face arrangement analysis. See the corresponding section in the Variables help topic for a list of all variables available.

The Face Arrangement Standard Category

The IMatch Standard Categories set is automatically imported when a new database is created. It contains a category named Face Arrangement. This is a data-driven category based on the variable {File.Faces.Arrangement.Type} which organizes your files based on the type determined by the face arrangement analysis. For example:

You can use this category directly, like any other category. Or you it with the Category Filter in the Filter Panel to quickly filter files by one of the groups.

If you don't see this category, just re-import the Standard Category set.
Open the Import & Export Panel, click on Category Import and import the file "C:\ProgramData\photools.com\IMatch6\Presets\system-en.imcsx".
If you prefer the German language set, import "system-de.imcsx" and for the set in Dutch, import "system-nl.imcsx" instead.

More Details

The standard category is simple by intention and considers only the Type. If you want to classify your files in more detail, add additional levels using one or more of the other {File.Faces.Arrangement} variables.

For the most detailed (and rarely needed) categorization, you can create a data-driven category based on {File.Faces.Arrangement} with the option Detect Hierarchies enabled and using ~; as the hierarch separator.

Face Arrangement Options

The default values work well for the majority of typical photographic scenarios. Your requirements may differ, though. Maybe you consider smaller faces also as portraits or your definition of close-up or small face differs. You can use the options described below to control the results of the face arrangement analysis performed by IMatch.

Go to Edit menu > Preferences > Application. Search for face and make sure show extended options at the bottom of the dialog is enabled.

IMatch caches the result of the face analysis in the database for fast access. When you change any of these settings, IMatch has to re-analyze all images containing at least one face. This takes only a few seconds per 100,000 images, depending on the performance of your computer.


Minimal face size for portrait

The minimal size of a face as a percentage of the image width / height to consider the face a portrait. Values around 30% are a good starting point.

Minimal face size for close-up

The minimal size of a face as a percentage of the image width / height to consider the face a portrait. Values around 70% are a good starting point.

See the following series of images for a better understanding of these two settings:

Small face size

Faces smaller than this value are considered small. Percentage of the image width / height. The smallest face in the file is used to determine the result. Values around 5% are a good starting point.

Group size

How many faces make a group? Only faces within Maximal group distance to each other are counted.

This value must be larger than 2 and less than Large group size.

Large group size

How many faces make a large group? Only faces within Maximal group distance to each other are counted.

This value must be larger than Group size.

Maximal group distance

The maximal distance in percent of the image width/height between two faces to include them into a group. Values around 15% are a good starting point.

Maximal close together distance

The maximal distance in percent of the image width/height between two faces to consider them as close together. Values around 15% are a good starting point.

IMatch excludes ignored faces when analyzing face arrangements.

Face annotations for Linked Persons have no real position or dimension. IMatch only counts them when analyzing face arrangements, but does not consider them when checking for groups or portraits/close-ups.