Free white paper
FaceReader Methodology
Learn about the science behind FaceReader. This white paper explains the methodology used for automatic facial expression analysis, including the classification of basic emotions, Action Units, and custom expressions.
Download for free
How does FaceReader work?
FaceReader was trained to classify the six basic or universal emotions: happy, sad, angry, surprised, scared, and disgusted. Additionally, FaceReader can recognize a neutral state.
The software uses a deep learning model (Deep Face) that was trained on over 10,000 manually annotated images. This model analyzes the face in three steps: face finding, face modeling, and face classification.
This white paper describes the methodology in detail, including validation studies and accuracy benchmarks.
In this white paper you will learn:
- checkHow the Active Appearance Model works
- checkHow Deep Face classifies expressions
- checkValidation and accuracy of FaceReader
- checkAction Unit classification
- checkCustom expression training