FERA 2017 - Addressing Head Pose in the Third Facial Expression Recognition and Analysis Challenge

Area of Science:

• Computer Vision
• Machine Learning
• Human-Computer Interaction

Background:

• Automatic Facial Expression Analysis (AFEA) has advanced significantly.
• Current AFEA evaluations often lack ecological validity, focusing on posed expressions or near-frontal views.
• This limits understanding of AFEA performance in real-world, varied-pose scenarios.

Purpose of the Study:

• To address the lack of suitable data for evaluating AFEA under diverse conditions.
• To present the third Facial Expression Recognition and Analysis (FERA 2017) challenge.
• To extend previous challenges by focusing on Action Unit (AU) occurrence and intensity estimation across different camera views.

Main Methods:

• Organized the FERA 2017 challenge, a sub-event of the 12th IEEE Conference on Face and Gesture Recognition.
• Defined two sub-challenges: detection of AU occurrence and estimation of AU intensity.
• Outlined the evaluation protocol and described the dataset used for the challenge.

Main Results:

• Presented results from a baseline method applied to both sub-challenges.
• Established a benchmark for evaluating facial expression recognition under varied poses and ecologically valid expressions.

Conclusions:

• The FERA 2017 challenge provides crucial resources for advancing AFEA research.
• The challenge facilitates the development of more robust and generalizable facial expression recognition systems.
• Future work will benefit from the established protocols and datasets for assessing performance in realistic scenarios.