Knowledge Base

Default Analysis Settings

Active Face Model

Face Model

Choose the model that will be used by FaceReader for finding, modeling and classifying faces. Note that for a specific analysis you can use another model. If you have Baby FaceReader there is only one face model available ('Baby').

See Settings For The Current Analysis in Analyze Facial Expressions

You can choose between the following models:

• General (default) — This model has been trained on a very diverse selection of images. The model works well under most circumstances for most people.
• Baby — Select this model if you work with babies or infants between the age of 6 months and 2 years. If you have a regular FaceReader license, the baby face model is not available.
• EastAsian — Select this model to analyze East Asian faces, for instance, Chinese or Japanese faces.

Calibration

Continuous Calibration

With this option selected, FaceReader continuously calibrates during an analysis.

See Continuous Calibration in Analyze Facial Expressions

Classification

Smoothen Classifications

With this option selected (default), classification values are transformed to obtain a smooth output signal. FaceReader smoothens the classification values taking the duration between frames into account. When there is a large time gap between frames, FaceReader will consider the facial expressions detected in those frames as less correlated then when there is a small time gap and thus adapts the strength of the smoothing.

When this option is not selected, FaceReader gives the raw output data.

If you are analyzing a set of images, this option has no effect.

Rotation

Image Rotation

If you have images that are rotated, select one of the options from the drop-down list to make sure the faces are upright.

• None (default).
• 90º CW (clockwise).
• 180º CW.
• 270º CW.

Video

Sample Rate

This option applies when you perform video analysis.

By default, Every frame is selected. This means that all video frames are analyzed. To speed up the analysis, change the sample rate to analyze Every 2nd frame or Every 3rd frame.

Under normal circumstances, analyzing 10–15 frames per second is sufficient for an accurate analysis. This means that if your video contains 30 frames per second or more, you can safely decrease the sample rate of FaceReader without loss of accuracy and make the analysis two to three times faster.

For camera analysis — Video is sampled at 15 frames per second, also when your camera is set to a higher frame rate, and independent of what you choose under Sample rate.

Source: FaceReader 10 Reference Manual (Help), Noldus Information Technology