Contrastive learning based remote sensing text-to-image generation for few-shot remote sensing image captioning

In few-shot scenarios, the lack of caption-labeled samples and prior knowledge leads to insufficient training and performance degradation of remote sensing image captioning (RC) models. We propose an iterative remote sensing image captioning method named IRIC to promote RC model performance iteratio...

Full description

Saved in:
Bibliographic Details
Main Authors: Haonan Zhou, Hang Tang, Xiangchun Liu, Xiaoxiao Shi, Lurui Xia
Format: Article
Language:English
Published: Taylor & Francis Group 2025-08-01
Series:International Journal of Digital Earth
Subjects:
Online Access:https://www.tandfonline.com/doi/10.1080/17538947.2025.2526102
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849224357312724992
author Haonan Zhou
Hang Tang
Xiangchun Liu
Xiaoxiao Shi
Lurui Xia
author_facet Haonan Zhou
Hang Tang
Xiangchun Liu
Xiaoxiao Shi
Lurui Xia
author_sort Haonan Zhou
collection DOAJ
description In few-shot scenarios, the lack of caption-labeled samples and prior knowledge leads to insufficient training and performance degradation of remote sensing image captioning (RC) models. We propose an iterative remote sensing image captioning method named IRIC to promote RC model performance iteration and generate higher quality captions. The IRIC first constructs a remote sensing text-to-image model CRTI based on contrastive learning, which can generate remote sensing images with the same semantic content from text and achieve text-driven remote sensing image transformation; Subsequently, caption-labeled sample amplification with prior knowledge introduction is implemented, which incorporates prior knowledge into the text-driven remote sensing image transformation to achieve caption-labeled sample amplification; Finally, the amplified caption-labeled samples are added to the original train set, and the RC model is retrained to achieve iterative performance improvement. The experimental results show that the IRIC is highly effective in few-shot scenarios and can iteratively improve the CIDEr scores of the latest few-shot RC model by 8.5%.
format Article
id doaj-art-4b80a9b4756c4a8d9706f854e062efd1
institution Kabale University
issn 1753-8947
1753-8955
language English
publishDate 2025-08-01
publisher Taylor & Francis Group
record_format Article
series International Journal of Digital Earth
spelling doaj-art-4b80a9b4756c4a8d9706f854e062efd12025-08-25T11:25:06ZengTaylor & Francis GroupInternational Journal of Digital Earth1753-89471753-89552025-08-0118110.1080/17538947.2025.2526102Contrastive learning based remote sensing text-to-image generation for few-shot remote sensing image captioningHaonan Zhou0Hang Tang1Xiangchun Liu2Xiaoxiao Shi3Lurui Xia4Beijing Institute of Remote Sensing Information, Beijing, People’s Republic of ChinaBeijing Institute of Remote Sensing Information, Beijing, People’s Republic of ChinaBeijing Institute of Remote Sensing Information, Beijing, People’s Republic of ChinaBeijing Institute of Remote Sensing Information, Beijing, People’s Republic of ChinaSpace Engineering University, Beijing, People’s Republic of ChinaIn few-shot scenarios, the lack of caption-labeled samples and prior knowledge leads to insufficient training and performance degradation of remote sensing image captioning (RC) models. We propose an iterative remote sensing image captioning method named IRIC to promote RC model performance iteration and generate higher quality captions. The IRIC first constructs a remote sensing text-to-image model CRTI based on contrastive learning, which can generate remote sensing images with the same semantic content from text and achieve text-driven remote sensing image transformation; Subsequently, caption-labeled sample amplification with prior knowledge introduction is implemented, which incorporates prior knowledge into the text-driven remote sensing image transformation to achieve caption-labeled sample amplification; Finally, the amplified caption-labeled samples are added to the original train set, and the RC model is retrained to achieve iterative performance improvement. The experimental results show that the IRIC is highly effective in few-shot scenarios and can iteratively improve the CIDEr scores of the latest few-shot RC model by 8.5%.https://www.tandfonline.com/doi/10.1080/17538947.2025.2526102Remote sensing image captioningremote sensing text-to-image generationcontrastive learningcaption-labeled sample amplification
spellingShingle Haonan Zhou
Hang Tang
Xiangchun Liu
Xiaoxiao Shi
Lurui Xia
Contrastive learning based remote sensing text-to-image generation for few-shot remote sensing image captioning
International Journal of Digital Earth
Remote sensing image captioning
remote sensing text-to-image generation
contrastive learning
caption-labeled sample amplification
title Contrastive learning based remote sensing text-to-image generation for few-shot remote sensing image captioning
title_full Contrastive learning based remote sensing text-to-image generation for few-shot remote sensing image captioning
title_fullStr Contrastive learning based remote sensing text-to-image generation for few-shot remote sensing image captioning
title_full_unstemmed Contrastive learning based remote sensing text-to-image generation for few-shot remote sensing image captioning
title_short Contrastive learning based remote sensing text-to-image generation for few-shot remote sensing image captioning
title_sort contrastive learning based remote sensing text to image generation for few shot remote sensing image captioning
topic Remote sensing image captioning
remote sensing text-to-image generation
contrastive learning
caption-labeled sample amplification
url https://www.tandfonline.com/doi/10.1080/17538947.2025.2526102
work_keys_str_mv AT haonanzhou contrastivelearningbasedremotesensingtexttoimagegenerationforfewshotremotesensingimagecaptioning
AT hangtang contrastivelearningbasedremotesensingtexttoimagegenerationforfewshotremotesensingimagecaptioning
AT xiangchunliu contrastivelearningbasedremotesensingtexttoimagegenerationforfewshotremotesensingimagecaptioning
AT xiaoxiaoshi contrastivelearningbasedremotesensingtexttoimagegenerationforfewshotremotesensingimagecaptioning
AT luruixia contrastivelearningbasedremotesensingtexttoimagegenerationforfewshotremotesensingimagecaptioning