Published on Thu Mar 05 2020

Unique Class Group Based Multi-Label Balancing Optimizer for Action Unit Detection

Ines Rieger, Jaspar Pahl, Dominik Seuss

Balancing methods for single-label data cannot be applied to multi-label problems. We propose to reformulate this problem as an optimization problem. We apply this balancing algorithm to training datasets for detecting isolated facial movements, so-called Action Units.

0
0
0
Abstract

Balancing methods for single-label data cannot be applied to multi-label problems as they would also resample the samples with high occurrences. We propose to reformulate this problem as an optimization problem in order to balance multi-label data. We apply this balancing algorithm to training datasets for detecting isolated facial movements, so-called Action Units. Several Action Units can describe combined emotions or physical states such as pain. As datasets in this area are limited and mostly imbalanced, we show how optimized balancing and then augmentation can improve Action Unit detection. At the IEEE Conference on Face and Gesture Recognition 2020, we ranked third in the Affective Behavior Analysis in-the-wild (ABAW) challenge for the Action Unit detection task.

Sat Feb 08 2020
Machine Learning
Multi-Label Class Balancing Algorithm for Action Unit Detection
Isolated facial movements, so-called Action Units, can describe combined emotions or physical states such as pain. As datasets are limited and mostly imbalanced, we present an approach incorporating a multi-label class balancing algorithm.
0
0
0
Mon Aug 17 2020
Machine Learning
Multi-label Learning with Missing Values using Combined Facial Action Unit Datasets
Facial action units allow an objective, standardized description of facial movements which can be used to describe emotions in human faces. Every study annotates different action units, leading to a tremendous amount of missing labels in a combined database. Our approach shows competitive performance compared to recent competitions in action unit detection.
0
0
0
Mon Apr 29 2019
Computer Vision
A Study on Action Detection in the Wild
Most common categories contain tens of thousands of examples, whereas rare ones have only dozens. We demonstrate that the standard AP metric is not informative for the categories in the tail. We propose an alternative one - Sampled AP.
0
0
0
Wed Mar 29 2017
Computer Vision
Learning with Privileged Information for Multi-Label Classification
A novel approach for learning multi-label classifiers with the help of privileged information. By integrating similarity constraints and ranking constraints into the learning process of classifiers, the privileged information and the dependencies among multiple labels are exploited.
0
0
0
Mon Feb 10 2020
Computer Vision
Multitask Emotion Recognition with Incomplete Labels
We train a unified model to perform three tasks: facial action unit. classification, expression classification, and valence-arousal estimation. We address two main challenges of learning the three tasks. Most existing datasets do not contain labeling for all three tasks, so we apply data balancing techniques.
0
0
0
Sun Jul 19 2020
Machine Learning
Distribution-Balanced Loss for Multi-Label Classification in Long-Tailed Datasets
Distribution-Balanced Loss is a new loss function for multi-label recognition problems. It takes into account the impact caused by label co-occurrence and the dominance of negative labels.
0
0
0