RGB and multispectral Household Smallholder Crops imagery Dataset

This dataset contains high-resolution multispectral and RGB crop imagery acquired using UAV platforms to support semantic segmentation, crop identification, and precision agriculture research. The data were collected in rural agricultural areas of Colombia (Silvia, Cauca) and correspond to representative smallholder crops ("pancoger"), including fique, potato, strawberry and coffee. Images were captured using two sensor systems: RGB camera (DJI Matrice 300 RTK + Zenmuse P1 camera): high-resolution visible imagery for fique, potato, and strawberry crops Multispectral (DJI Mavic 3M): multispectral and RGB imagery for coffee crops, including 5-band master TIFF images (R, G, B, Red Edge, NIR) Each image tile includes a corresponding ground-truth segmentation mask, manually annotated following a consistent labeling protocol, and NDVI indexes for the multispectral coffee imagery. Masks and indices follow a 1:1 correspondence with each source image. The dataset is structured following FAIR principles to ensure reproducibility and interoperability. All images are accompanied by: A metadata.csv file describing each image (sensor, crop type, dimensions, masks, NDVI availability, and file formats) A data_dictionary.csv detailing all variables included in the metadata A comprehensive README explaining folder structure and processing workflows A LICENSE file specifying the dataset usage policy This dataset supports research in: Deep learning for semantic segmentation Crop classification Multispectral index analysis Agricultural monitoring in smallholder farming systems UAV-based remote sensing Researchers may use this dataset to train and evaluate models such as U-Net, DeepLab, Mask R-CNN, or transformer-based architectures applied to crop segmentation and multispectral analysis.