msuav500k: Foundational dataset for multispectral and RGB uncrewed aerial vehicle imagery

This dataset comprises a curated collection of UAV RGB and multispectral imagery sourced from multiple open-access platforms. It contains the largest set of UAV imagery covering at least four spectral bands (Green, Red, RedEdge, and Near-Infrared) and is available as either orthomosaic or raw sensor data. A systematic calibration process was employed to ensure radiometric consistency across the diverse sensor types (DJI Mavic 3 M, DJI Phantom 4 Multispectral, Parrot Sequoia, MicaSense RedEdge, and MicaSense Altum(PT)). This involved correcting raw digital numbers (DNs) for sensor-specific variables (e.g., black level, vignetting) and normalizing pixel values with internal sunlight sensors. The resultant imagery is provided as UINT8 5-channel multispectral TIFF files and RGB 3-channel JPG files for robust cross-study comparison. Comprehensive metadata, processing scripts, and calibration details are publicly available in the accompanying repository folder. This dataset offers a valuable resource for researchers and practitioners seeking consistent, high-quality UAV multispectral data to train or fine-tune foundational models in computer vision and remote sensing applications.