MonoIR-RS: Infrared Remote Sensing Vision-Language Learning with CLIP-Style Contrastive Adaptation

Forum topic · 小凯 · 2026-07-09

Summary

MonoIR-RS (arXiv:2507.06827) is a large-scale infrared remote-sensing vision-language dataset and benchmark addressing the underexplored domain of infrared vision-language understanding. Built from the same source pool and split as FusionRS, it retains infrared imagery as the model-facing modality, producing 600,000 synthesized infrared images and 59,032 IR-aware caption records whose supervision emphasizes grayscale structure and IR-style contrast rather than RGB appearance. The authors demonstrate that their synthesized infrared images approximate real thermal imagery noticeably better than grayscale conversion on the AVIID benchmark. They fine-tune five CLIP backbones and six VLM backbones, showing that IR-aware adaptation improves CLIP average recall by up to 12.8 points, drives infrared cue coverage in VLM captions to 100%, and reduces residual RGB color leakage to near zero. By isolating the infrared modality from RGB-IR multimodal learning, MonoIR-RS provides a controlled, reproducible testbed for aligning infrared remote-sensing evidence with language.

Paper Overview

Field: Computer Vision Authors: Jiaju Han, Yaqi Ma, Yahui Chai Published: 2025-07-09 arXiv: 2507.06827

Abstract

Infrared remote-sensing imagery captures intensity structure, object-background contrast, and illumination-invariant cues often invisible in RGB imagery. Yet, most remote-sensing vision-language resources and models focus on visible-band semantics, leaving infrared vision-language understanding underexplored. The authors introduce MonoIR-RS, a large-scale infrared remote-sensing vision-language dataset and benchmark that couples IR-aware data construction with CLIP-style contrastive adaptation and VLM instruction tuning.

Built from the same source pool and split as FusionRS, MonoIR-RS retains the infrared image as the model-facing modality, yielding 600,000 synthesized infrared images and 59,032 retained IR-aware caption records. The model experiments use this retained language-supervision subset, whose captions are rewritten to supervise grayscale structure and infrared-style contrast rather than RGB appearance.

Key Findings

Synthesized infrared images are shown to approximate real thermal imagery noticeably better than grayscale conversion on the AVIID benchmark.
Five CLIP backbones and six VLM backbones were fine-tuned and compared against zero-shot baselines.
IR-aware adaptation improves CLIP average recall by up to 12.8 points.
IR cue coverage in VLM-generated captions reaches 100%, while residual RGB color leakage is reduced to near zero.

By isolating the infrared modality from RGB-IR multimodal learning, MonoIR-RS provides a controlled, reproducible testbed for aligning infrared remote-sensing evidence with language.