TIR-RGB-UAV: A Multimodal, High-Resolution, and Multitemporal Dataset for Urban Scene Analysis

The dataset used consists of multispectral RGB and TIR images acquired using a Unmanned Aerial Vehicle (UAV), commonly referred to as a drone, equipped with ultra-high-resolution sensors. Data collection was conducted over 12 months, with one flight campaign per month. For each campaign, flights were carried out during three distinct times of the day: morning, afternoon, and night. On four occasions, nighttime flights were not performed due to unfavorable weather conditions, such as rainfall and strong winds, which made UAV operations unsafe.A total of 13 land use and land cover classes were defined, with four samples collected per class during each flight. As a result, the dataset comprises 1,664 RGB images and 1,664 TIR images, acquired simultaneously by the dual-sensor camera onboard the RPA, ensuring temporal alignment between spectral domains.It is important to note that due to the optical characteristics of the sensors, the RGB and TIR images differ in both spatial resolution and field of view. The ground sampling distance (GSD) for the RGB images is approximately 4 cm, while for the TIR images it is around 18 cm. Furthermore, the RGB images capture a slightly larger area compared to the TIR images, requiring pre-processing steps such as spatial registration and cropping to enable proper spectral fusion.