Efficient Remote Sensing with Harmonized Transfer Learning and Modality Alignment

arXiv:2404.18253v1 Announce Type: cross
Abstract: With the rise of Visual and Language Pretraining (VLP), an increasing number of downstream tasks are adopting the paradigm of pretraining followed by fine-tuning. Although this paradigm has demonstrated potential in various multimodal downstream tasks, its implementation in the remote sensing domain encounters some obstacles. Specifically, the tendency for same-modality embeddings to cluster together impedes efficient transfer learning. To tackle this issue, we review the aim of multimodal transfer learning for downstream tasks from a …

alignment arxiv cs.cv cs.lg sensing transfer transfer learning type