Related Experiment Video
Updated: Jan 28, 2026

Author Spotlight: Addressing Technical and Subjective Challenges in Measuring Classroom Attention
Published on: December 15, 2023
LLM-Based Pose Normalization and Multimodal Fusion for Facial Expression Recognition in Extreme Poses
Bohan Chen1, Bowen Qu1, Yu Zhou1
1School of Information Engineering, Zhongnan University of Economics and Law, Wuhan 430073, China.
This study introduces a novel facial expression recognition (FER) method that transforms profile images to frontal views and uses multimodal learning. This approach significantly improves accuracy for non-frontal facial images.
Area of Science:
- Computer Science
- Artificial Intelligence
Background:
- Facial expression recognition (FER) systems struggle with accuracy for non-frontal or rotated faces.
- Current FER methods are limited by pose variations, hindering real-world applications.
Purpose of the Study:
- To enhance FER accuracy by addressing challenges posed by large pose variations.
- To develop a robust FER method capable of recognizing expressions from profile and large-angle rotated facial images.
Main Methods:
- Proposed a novel FER method combining profile-to-frontal image transformation and multimodal learning.
- Utilized Qwen-Image-Edit for transforming profile images to frontal views, preserving expression features.
- Employed the CLIP model for vision-language joint learning to improve semantic representation of expression features.
Main Results:
- Achieved high accuracy on benchmark datasets: RAF (89.39%), EXPW (67.17%), and AffectNet-7 (62.66%).
- Demonstrated superior performance compared to existing FER approaches, particularly for challenging poses.
Conclusions:
- The proposed profile-to-frontal transformation and multimodal learning approach effectively mitigates pose variation issues in FER.
- This method offers a significant advancement for practical and accurate facial expression recognition systems.
More Related Videos
05:51Exploring the Use of Isolated Expressions and Film Clips to Evaluate Emotion Recognition by People with Traumatic Brain Injury
Published on: May 15, 2016
05:38Interaction between Phonological and Semantic Processes in Visual Word Recognition using Electrophysiology
Published on: June 29, 2021
Related Concept Videos
Muscles for Facial Expressions
Nuclear Fusion
A helium nucleus has a mass that is 0.7% less than that of four hydrogen nuclei; this lost mass is converted into energy during the fusion. This reaction produces about...
Facial Feedback Hypothesis
Normal Stress
When a rod is under axial loading, the internal forces and corresponding stress are normal to the plane of the section, so it is termed normal stress. It's important to...
Absolute and Local Extreme Values
Normal Distribution