Increasingly Packing Multiple Facial-Informatics Modules in A Unified Deep-Learning Model via Lifelong Learning

Simultaneously running multiple modules is a key requirement for a smart multimedia system for facial applications including face recognition, facial expression understanding, and gender identification. To effectively integrate them, a continual learning approach to learn new tasks without forgetting is introduced. Unlike previous methods growing monotonically in size, our approach maintains the compactness in continual learning. The proposed packing-and-expanding method is effective and easy to implement, which can iteratively shrink and enlarge the model to integrate new functions. Our integrated multitask model can achieve similar accuracy with only 39.9% of the original size.

PDF Abstract

Results from the Paper

 Ranked #1 on Gender Prediction on FotW Gender (using extra training data)

     Get a GitHub badge
Task Dataset Model Metric Name Metric Value Global Rank Uses Extra
Training Data
Age And Gender Classification Adience Age PAENet (single crop, tensorflow) Accuracy (5-fold) 57.3 # 9
Age And Gender Classification Adience Gender PAENet (single crop, tensorflow) Accuracy (5-fold) 89.08 # 4
Facial Expression Recognition (FER) AffectNet PAENet Accuracy (7 emotion) 65.29 # 10
Accuracy (8 emotion) - # 27
Continual Learning Cifar100 (20 tasks) PAENet Average Accuracy 77.1 # 6
Gender Prediction FotW Gender PAENet Accuracy (%) 92.93 # 1


No methods listed for this paper. Add relevant methods here