Deep domain adaptation with ordinal regression for pain assessment using weakly-labeled videos

作者：

Highlights：

• Pain assessment from facial expressions is addressed using weakly-labeled videos.

• An adversarial method for deep domain adaptation through multiple instance learning

• Ordinal relations between pain intensity levels is modeled with Gaussian distribution.

• Pooling strategy is improved by adaptively selecting frames relevant to a bag label.

• Extensive experiments show that our proposed WSDA outperforms state-of-the-art models.

摘要

•Pain assessment from facial expressions is addressed using weakly-labeled videos.•An adversarial method for deep domain adaptation through multiple instance learning•Ordinal relations between pain intensity levels is modeled with Gaussian distribution.•Pooling strategy is improved by adaptively selecting frames relevant to a bag label.•Extensive experiments show that our proposed WSDA outperforms state-of-the-art models.

论文关键词：Deep domain adaptation,Weakly-supervised learning,Multiple instance learning,Ordinal regression,Pain intensity estimation

论文评审过程：Received 3 November 2020, Revised 1 March 2021, Accepted 4 April 2021, Available online 8 April 2021, Version of Record 21 April 2021.

论文官网地址：https://doi.org/10.1016/j.imavis.2021.104167