Detecting Faces
Applies to: Face Detection and Recognition
Face detection and recognition was introduced to Timestone Software with version 4.10 of its applications. This document discusses the key points of face detection.
Installing the face tools
It is necessary to install the Face Tools toolkit on each computer where face detection and recognition is to be performed. The toolkit can be downloaded from the Timestone Software download pages.
Before you begin
To successfully detect faces in images, consideration should be given to the following.
Accessibility of the original images
Face detection is performed on the original images not the image previews. It is therefore imperative that image paths are correct and the software can find the original images. To this end running the 'check images' report (Images Mode>Images>Check ..) can be useful as it will alert the user of missing images.
Image preview orientation
Faces need to be in the correct, upright, orientation in order to be detected. Portrait images that have been imported in the landscape orientation should therefore be rotated before face detection.
Filtering and selecting images
Face detection occurs on the selected images only. The user may therefore wish to use the various Images Mode filters before selecting the images for face detection. This avoids performing the face detection on images where it is not needed.
Show lines
A successfully detected face will display a teal or red coloured box around the face. This box can be toggled on and off with the menu option Images Mode>Images>Show face lines or with the Ctrl-Alt-F5 keyboard shortcut.
Detecting faces
With the images selected, detecting faces is simply a matter of using the keyboard shortcut, Ctrl-F5. The same action can also be achieved via the menu option Images Mode>Images>Recognize faces...
A progress indicator appears and as each face is detected a teal coloured box appears for each face. A red coloured box indicates closed eyes.
Figure 1: Images after Ctrl-F5 to detect faces.
Note the last two images in the screen grab above. Since they are not in the correct orientation, the face detection was not successful. The remaining images all display a teal coloured box around the face, indicating the face was detected and eyes are open.
Checking for no faces
A 'no faces' filter in Images Mode allows the user to quickly and easily find any images where faces have not been detected.
If no faces are detected in any images
Assuming the face tools were installed on the computer and face detection has worked before, check the following:
- Ensure the software can find the original images. Run the check images report
- Check the images were actually selected and they are in the correct orientation
- Check Show face lines is selected
When some images do not have a face detected
There are some circumstances where the software cannot detect a face in the image even though all the above criteria have been met. These include
- the face is tilted too far resulting in the eyes being too far away from horizontal
- only one side of the face is visible
- one side of the face is in deep shadow
Figure 2: Using the no faces filter after the face detection
Face detection in CapturePost
Unlike the other apps, face detection in CapturePost occurs auomatically as the images are captured and saved.
Face detection in NeoGroup
Face detection in NeoGroup enables the user to automatically name a group photograph. It is discussed here
Naming groups with face detection and recognition
End of article