Learning with Noisy Labels for Human Fall Events Classification: Joint Cooperative Training with Trinity Networks

TL;DR

This paper introduces JoCoT, a novel training framework using Trinity Networks and consensus-based guidance to improve human fall event classification accuracy under noisy labels while preserving privacy.

Contribution

The paper proposes JoCoT, a new noisy label learning method employing Trinity Networks and privacy-preserving skeleton data for improved fall detection accuracy.

Findings

JoCoT outperforms state-of-the-art methods by 5.17% and 3.35% under high noise conditions.

The framework effectively handles noisy labels in fall event classification.

Consensus-based teacher guidance enhances model robustness.

Abstract

With the increasing ageing population, fall events classification has drawn much research attention. In the development of deep learning, the quality of data labels is crucial. Most of the datasets are labelled automatically or semi-automatically, and the samples may be mislabeled, which constrains the performance of Deep Neural Networks (DNNs). Recent research on noisy label learning confirms that neural networks first focus on the clean and simple instances and then follow the noisy and hard instances in the training stage. To address the learning with noisy label problem and protect the human subjects' privacy, we propose a simple but effective approach named Joint Cooperative training with Trinity Networks (JoCoT). To mitigate the privacy issue, human skeleton data are used. The robustness and performance of the noisy label learning framework is improved by using the two teacher modules and one student module in the proposed JoCoT. To mitigate the incorrect selections, the predictions from the teacher modules are applied with the consensus-based method to guide the student module training. The performance evaluation on the widely used UP-Fall dataset and comparison with the state-of-the-art, confirms the effectiveness of the proposed JoCoT in high noise rates. Precisely, JoCoT outperforms the state-of-the-art by 5.17% and 3.35% with the averaged pairflip and symmetric noises, respectively.