A Multi-Task Framework for Facial Attributes Classification through End-to-End Face Parsing and Deep Convolutional...

Area of Science:

• Computer Vision
• Machine Learning
• Biometrics

Background:

• Human face image analysis is a challenging computer vision task.
• Accurate recognition of demographic attributes like race, age, and gender from faces remains difficult.

Purpose of the Study:

• To propose a novel framework for face image analysis.
• To address the challenges of race, age, and gender recognition using face parsing.
• To develop an end-to-end deep learning model for face parsing.

Main Methods:

• Manually labeled face images were used to train a Deep Convolutional Neural Network (DCNN) for face parsing.
• The DCNN model segments face images into seven dense classes.
• Probabilistic classification generated probability maps for each face class, serving as feature descriptors.
• A separate DCNN model was trained using these probability maps for demographic recognition tasks.

Main Results:

• The proposed face parsing framework achieved significantly improved results on state-of-the-art datasets.
• The probability maps derived from face parsing proved effective as feature descriptors for demographic recognition.
• The system demonstrated superior performance in race, age, and gender recognition compared to previous methods.

Conclusions:

• The developed framework offers a robust approach to human face image analysis.
• Face parsing combined with probability maps is a promising technique for demographic attribute recognition.
• This research advances the state-of-the-art in computer vision for face analysis.