What are filters and how do they work

Facial recognition software can be quite sensitive so we've included a number of tools to help make sure that your captions only contain people who are relevant.

The currently available filters are below. We will discuss the difference between pre and post recognition filters and each filter in more depth below.

Pre recognition filters:  

  • Max faces per image

  • Only use local database

Post recognition filters:

  • Face size

  • Face sharpness

  • Solo people

  • Enabled guests

How to edit filters

Defaults for filters are set in the project settings window when creating a new project, or editing an existing project. You can also change a selection of filters on an image by image basis from the filters section in the right hand part of the screen once an image is selected. You can edit filters for multiple images at the same time by selecting multiple images and then editing the filters. Editing the project filter defaults will only apply to images that are processed after the defaults have been changed.

How to tell if a filter is active

There are two ways to tell if a filter is being applied to a face. Firstly the face box in the image will become more transparent and less pronounced. Secondly, when that face is selected, the currently active filters will be outlined in red.

What's the difference between pre and post recognition filters?

In simple terms pre recognition filters require an image reprocess to be changed, whereas post recognition filters do not. Its best to set pre recognition filters once for the project so that they do not have to be changed, whereas post recognition filters should be set with sensible defaults, and tweaked on each image.

The filters

Max faces per image (pre recognition) 

This sets a limit on the number of faces that Caption Pro will detect in the image. There are two main reasons you may want to lower this value. 

Firstly, to limit your monthly costs. Caption Pro charges based on the number of faces detected in an image, if for instance you are a sports photographer and Caption Pro detects 30 faces in the crowd of a football game when all you are interested in are the players on the field, you will be paying for more faces than is necessary.

Secondly, to improve processing speed. Images are faster to process if less faces are detected.

We recommend setting this value slightly higher than the maximum number of people of interest that are likely to appear in an image. This covers all eventualities and avoids the time consuming process of having to change the filter and reprocess the images containing more faces than you anticipated.

This filter will prioritize faces that appear larger in the image.

Only use local database (pre recognition) 

This will ensure Caption Pro only suggests people that have been added to your database by you or your team. It will not recommend new previously unseen celebrities from Caption Pro's celebrity database. There are two reasons why you may want to use this. 

Firstly, to improve processing speed. Looking up people in Caption Pro's celebrity database is the slowest part of image processing, and on large images with many faces and a high "max faces per image" set, this can become very slow. 

Secondly, to prevent unwanted name suggestions. In certain situations you may already have everybody you know who is going to be at an event in your facial recognition database, in which case any new suggestion is likely to be incorrect and will slow down your captioning process.

Face size

The face size filter allows you to filter out smaller faces from your image. This can be changed by using the relevant slider in the filter section of the UI, or by using the hotkeys: - and =.

Face sharpness

The face sharpness filter filters out people based on how blurry their face is. This can also be changed by using the relevant slider in the filter section of section of the UI, or by using the hotkeys: [ and ]

Guests enabled

By toggling guests enabled faces that have not been recognized will be labelled as "Guest". Toggle guests enabled in the filter section of the UI or by using the hotkey g

Solo people

Caption Pro will ignore all people in the image apart from those in the "people to solo" input box. If no names have been input into the box this filter has no effect. This can be particularly useful when you know a group of images only contain certain people, as you may edit this field (and other fields) with multiple images selected. Please note that you may only add people to this box if they have been previously recognized in any of your projects.

Manually adding / removing a face 

Adding a face

Sometimes Caption Pro is unable to detect a face in an image even though it is there. This can be due to multiple reasons, such as the face being too blurry, the person looking away, or the face being obscured. However in these cases it is still possible to quickly add the subject's name to the caption. You can do this by right clicking the correct location in the image and selecting 'add face'. It is important that this is done in the correct location so that the name appears in the right order within the caption. 

After you have added a face box Caption Pro will allow you to associate a person with this face by searching your local database, or adding a new person record. User added faces will not become reference faces as Caption Pro doesn't have any facial recognition data for them. 

Manually added face boxes won't count as being processed and so will not be charged.

Removing a face

Deleting a face box in an image works the same way whether you manually added it, or Caption Pro detected it automatically. It will not delete the associated person from the database. It will simply ignore that face in the image in question, removing that person's name from the list of names in the currently selected image.

To delete a face right click on the face box in the image and select "delete face" from the context menu.

What's next?

Next, lets look at how to Edit Metadata


Did this answer your question?