CLIP-Guided Dynamic Feature Fusion With Relation-Enhanced Attention for Remote Sensing Image Captioning

Remote sensing image captioning is a multimodal task that aims to automatically generate descriptions for remote sensing images. However, the inherent scale diversity and semantic complexity of RSI pose significant challenges in semantic alignment and explicit modeling of complex scene structures. T...

Ausführliche Beschreibung

Bibliografische Detailangaben

Hauptverfasser:	Haifeng Sima, Wenqing Jiang, JianLong Wang, Changchun Li, Mingliang Xu
Format:	Artigo
Sprache:	Inglês
Veröffentlicht:	IEEE 2026-01-01
Schriftenreihe:	IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
Schlagworte:	Contrastive language–image pretraining (CLIP) fine-tuning dynamic feature fusion relation-enhanced attention (REA) remote sensing image captioning (RSIC) semantic alignment
Online-Zugang:	https://ieeexplore.ieee.org/document/11488868/
Tags:	Tag hinzufügen Keine Tags, Fügen Sie das erste Tag hinzu!