Source-Free Domain Adaptation Guided by Vision and Vision-Language Pre-Training

arXiv:2405.02954v1 Announce Type: cross
Abstract: Source-free domain adaptation (SFDA) aims to adapt a source model trained on a fully-labeled source domain to a related but unlabeled target domain. While the source model is a key avenue for acquiring target pseudolabels, the generated pseudolabels may exhibit source bias. In the conventional SFDA pipeline, a large data (e.g. ImageNet) pre-trained feature extractor is used to initialize the source model at the start of source training, and subsequently discarded. Despite having diverse features …

abstract adapt arxiv bias cs.cv cs.lg domain domain adaptation free generated key language pre-training training type vision vision-language while