Naming Groups using Face Recognition

Rather watch a video than read a document ?

A movie tutorial accompanies this article.

Scope

Faces within the group photo are detected then cross-referenced against the portrait faces of subjects in the same job. When a match is found, the subject details (usually their name) are automatically entered into the correct row and position of the name list.

When faces in the group photo are detected but not recognized (usually due to no portrait image for the subject) the user can perform a further lookup of other jobs that are part of the same job collection). They can also lookup the name of the unrecognized face or simply enter it manually - assuming in both cases they know who it is.

Before you begin

As per the detection of portrait images, NeoGroup requires access to the original group images. The images should also be in their correct orientation.

Although faces in the group photograph can be detected at any time, the recognition step requires the portrait image faces to have already been detected.

PREREQUISITES

• Face detection should be done on portrait images

• NeoGroup requires access to the original images - ensure image paths are correct

Note it is not necessary to perform face detection on the group images in Images Mode.

QUICK LINK

Further information on Detecting Faces can be found here:

Detecting Faces

Naming the groups in Names Mode

With the above criteria having been met, naming the groups is then simply a matter of:

1. Enter Names Mode (F6)

2. Select one or more groups

3. From the menu select Names>Detect and recognize faces (Ctrl-F5)

The software scans the group image, detecting each face - indicated by a white box. Once the detection phase is complete, it recognizes the face in the group photo against the portrait image faces detected in the other application. When a successful match is made, the white box turns green and the name is placed into the correct row and position in the names list.

Faces detected but not recognized

Occasionally, faces in the group photo are detected but not recognized against a matching portrait image. In these cases the box around the face remains white and a yellow alert symbol alerts the user. These situations are almost always due to the subject in question not having a portrait image. In these cases the user can enter the name manually, use the lookup tool to find the subject or perform a lookup of other jobs (notably the previous year’s job) via the use of job collections.

In the screen grab above, the teacher's face was detected but since she was not photographed there is no corresponding portrait image for the software to match to. Note the yellow alert symbol against her row and against the group name itself.

Lookup old jobs for unrecognized faces

Thanks to the use of Job Collection, it is possible for NeoGroup to search for unrecognized faces in other jobs. The lookup will search for the unrecognized faces in both the group photos and the subject portrait images provided face detection has been performed on the latter.

To identify unrecognized faces by searching in other jobs:

1. Create a collection that includes the current job and jobs to be searched

2. Ensure face detection has been performed on the portraits for jobs in the collection

3. Select Names>Identify unrecognized faces (Ctrl-Alt-F5)

4. From the popup window, confirm identified faces

QUICK LINK

More information on identifying unrecognized faces in groups is available here:

Identify unnamed group faces using past jobs

Faces not detected at all

Very rarely, a face in the group photography may be detected at all. This is more likely with very large groups such as year level or whole school group photos where faces are very small and the camera resolution is not very high. In these cases it is possible to do a more sensitive face detection by invoking Shift-Ctrl-F5. This performs a deeper, more sensitive scan that has been found to always find faces irrespective of their size or the size of the group image.

In the screen grab above, the face in the back row, 2nd from left was not detected. Repeating the step with Shift-Ctrl-F5 detects the face.

The more sensitive Shift-Ctrl-F5 increases the likelihood of false faces (artefacts). It is therefore strongly recommended it is only used for the largest of groups