IRNova/IRNova

IRNova是一个近红外（NIR）基准数据集，旨在研究计算机视觉任务（如图像分类和目标检测）中的跨光谱鲁棒性。该数据集包含使用我们提出的CRIG-Net框架从RGB图像生成的合成NIR图像，这是一种跨模态RGB到NIR生成策略，旨在适应光谱特征的同时保持结构一致性。该存储库目前提供用于评审评估和研究目的的基准子集。IRNova is a Near-Infrared (NIR) benchmark dataset designed for studying cross-spectral robustness in computer vision tasks such as image classification and object detection. The dataset contains synthetic NIR images generated from RGB images using our proposed CRIG-Net framework, a cross-modal RGB-to-NIR generation strategy designed to preserve structural consistency while adapting spectral characteristics. This repository currently provides benchmark subsets intended for reviewer evaluation and research purposes.