TeX-1500
收藏资源简介:
TeX-1500是由西湖大学研究团队构建的大规模配对长波红外高光谱成像数据集,旨在为温度-发射率-纹理分解任务提供监督学习基准。该数据集包含1,522个真实场景样本,整合了DARPA IH推扫式成像数据和自主采集的FTIR数据,覆盖不同地理位置、季节、采集时间及传感器类型,每个样本包含校准后的有效波段辐射立方体、波长位置及对齐的物理解释标签。数据集通过统一的预处理协议构建,包括去噪、波长校准和物理解释标签生成,专门用于推动基于学习的红外高光谱到物理属性分解研究,解决传统热感知中物理属性监督数据缺乏的关键问题。
TeX-1500 is a large-scale paired long-wave infrared hyperspectral imaging dataset constructed by the research team from Westlake University, designed to provide a supervised learning benchmark for the temperature-emissivity-texture decomposition task. This dataset includes 1,522 real-world scene samples, integrating DARPA IH push-broom imaging data and independently collected FTIR data, covering various geographical locations, seasons, acquisition times and sensor types. Each sample contains calibrated effective-band radiance cubes, wavelength positions, and aligned physical interpretation labels. The dataset is built through a unified preprocessing protocol, which includes denoising, wavelength calibration and physical interpretation label generation, and is specifically dedicated to advancing learning-based research on infrared hyperspectral-to-physical property decomposition, addressing the critical issue of insufficient supervised physical property data in traditional thermal perception.




