UniRGB-IR: A Unified Framework for RGB-IR Downstream Tasks

Summary:
- UniRGB-IR proposes a scalable and efficient framework to unify RGB and infrared (IR) image processing tasks using a Vision Transformer as the foundation.
- A Multi-modal Feature Pool (MFP) and a Supplementary Feature Injector (SFI) are used as adapters to enhance ViT features with RGB-IR contextual information.
Key Findings:
- The framework demonstrates state-of-the-art performance on various RGB-IR downstream tasks, achieving significant improvements in image analysis under low-light conditions.
- The uniqu…
Personalized AI news from scientific papers.