Improved multi-source domain adaptation by preservation of factors

作者：

Highlights：

• Introduction of visual factor theory to describe domains and interpret DA approaches

• Novel training method FP-DA allows to preserve a chosen factor during DA

• Use of PCA to show how it helps to find factors worth to be preserved

• Superior performance of FP-DA on two multi-domain object classification datasets

摘要

•Introduction of visual factor theory to describe domains and interpret DA approaches•Novel training method FP-DA allows to preserve a chosen factor during DA•Use of PCA to show how it helps to find factors worth to be preserved•Superior performance of FP-DA on two multi-domain object classification datasets

论文关键词：Domain adaptation,Multi-source domain adaptation,Adversarial domain adaptation,Negative transfer,Visual factors,Domain factors

论文评审过程：Received 17 September 2020, Revised 4 May 2021, Accepted 5 May 2021, Available online 8 May 2021, Version of Record 23 June 2021.

论文官网地址：https://doi.org/10.1016/j.imavis.2021.104209